This article introduces synthetically using RS, GIS and GPS (3S) technology to study ecosystem changes of Talimu river system within past 11 years. In this article, the data were analyzed by using GIS technology and the field sizes were investigated by using GPS instrument. A conclusion was concluded that in the past 11 years 1990 to 2000, in the region of the middle and lower reaches of Talimu river, the area of farmland had been increased 505,433km², the area of nature/planted vegetation had been reduced 943,089km², and the area of water body had been increased 80.477%. The area of desert in the middle and lower reaches of Tamilu river is increasing continually. The main reason is that the large number of nature vegetation was died, especially in the area of Huyang woods was declined. Finally, some suggestions were make to recover and father fragile ecosystem.

Single sensor night vision system has been developed for many years, but it still has many intrinsic limitations. For example, in a single channel imaging system the image contrast will be too low to distinguish different targets. As a new technique, multi-sensor image fusion color night vision is developed. This system can make up a single sensor's limitation and improve the system's performance. Depending on Sharmon's formula of the information theory, this paper describes the formula of the fusion image's amount of information and proves that multi-sensor image fusion technique is good at increasing amount of information. Then according to the information theory, the LLLCCD and IRFPA image fusion system is developed. The system's block diagram is given and described in detail in this paper. And the improved fusion algorithm is adopted which is especially good at this fusion system. At last the impersonal judgment of fusion effect is employed to analysis this system's performance and the practical image is given to show this system's good effect.

This paper is focused on the restoration of color remote sensing (including airborne photo). A complete approach is recommended. It propose that two main aspects should be concerned in restoring a remote sensing image, that are restoration of space information, restoration of photometric information. In this proposal, the restoration of space information can be performed by making the modulation transfer function (MTF) as degradation function, in which the MTF is obtained by measuring the edge curve of origin image. The restoration of photometric information can be performed by improved local maximum entropy algorithm. What's more, a valid approach in processing color remote sensing image is recommended. That is splits the color remote sensing image into three monochromatic images which corresponding three visible light bands and synthesizes the three images after being processed separately with psychological color vision restriction. Finally, three novel evaluation variables are obtained based on image restoration to evaluate the image restoration quality in space restoration quality and photometric restoration quality. An evaluation is provided at last.

Color image process is a very important problem. However, the main approach presently of them is to transfer RGB colour space into another colour space, such as HIS (Hue, Intensity and Saturation). YIQ, LUV and so on. Virutally, it may not be a valid way to process colour airborne image just in one colour space. Because the electromagnetic wave is physically altered in every wave band, while the color image is perceived based on psychology vision. Therefore, it's necessary to propose an approach accord with physical transformation and psychological perception. Then, an analysis on how to use relative colour spaces to process colour airborne photo is discussed and an application on how to tune the image tone in colour airborne image mosaic is introduced. As a practice, a complete approach to perform the mosaic on color airborne images via taking full advantage of relative color spaces is discussed in the application.

Given essential matrix, an analytical solution of the camera's motion parameteres is provided in this short note. To our knowledge, there has been no similar report in the literature up to now.

This paper presents a novel approach for creating 3D free-form scene models from a single paiting or photograph with no prior knowledge about the shape. The new technique takes as input a sparse set of user-specified constraints, and generates a well-behaved 3D surface satisfying the parameters. As each constraint is specified, the system recalculates and displays the reconstruction in real time. In contrast to previous work in single view reconstruction, our technique enables high quality reconstruction of free-form curved surfaces. A key feature of the approach is a novel hierarchical transformation for accelerating convergence on a non-uniform, piecewise continuous grid. The technique is interactive and updates the model in real time as constraints are added, allowing fast reconstruction of photorealistic scene models. The approach is shown to yield high quality results.

Due to the complex motion of a target, the Doppler frequency shifts are time-varying. Therefore, the inverse synthetic aperture radar (ISAR) image is blurred by using the Fourier analysis for Doppler processing. To resolve the image blurring problem, a parametric method which combines the matching Fourier transform with the CLEAN technique is presented to retrieve Doppler information. For comparison, a time-frequency transform, named reassigned Gabor spectrogram, is also applied to the time-varying spectral analysis. Simulation results show that the suggested method is robust to small data size and low signal-to-noise ratio.

This paper presents a wavelet-domain Hidden Markov Tree(HMT)-based color image superresolution algorithm using multi-channel data fusion. Because there exists correlations among the three channels of a RGB color image, a channel by channel superresolution method almost certain leads to color distortion. In order to solve this problem, first the low-resolution color image is converted into a gray-scale image using the spatially-adaptive approach presented in this paper and the resulting gray-scale image must reflect the human perception of edges in the color image; then by superresolving this gray-scale image, a high-resolution image is obtained; finally, wavelet-domain HMT-based image superresolutions are performed for the three channels of the low-resolution color image using the same posterior state probabilities, which reflect the hidden states of the wavelet coefficients of the high-resolution gray-scale image obtained before, and thus the resulting high-resolution color image is what we desired. Becasue the correlations among the three channels of a RGB color image are considered, there are no color distortions in the reconstructed high-resolution image. Experimental results show that the reconstructed color images have high PSNR and are of high visual quality.

A method for automatic segmentation and detection of small target in the earth to sky background was presented in this paper. When small target in far distance some objects on ground came upon to the scene. This always resulted in false detection in automatic detecting systems if we segmented small object by only using its illumination. In order to remove the big block and disconnected part of a gradient image was made because the gradient scale of the small target and background objects may be comparatively similar. Then an adaptive threshold method was adopted by using the image means as a threshold for several times to segment objects in the gradient image. After that successive over-relaxation was used to incorporate the disperse regions and the nearby isolated points would connect to the big block. The lowest value of valley (nonzero value) was searched and this value was used as the threshold to make binary image. So the connected clutters could be removed only by counting the number of pixels in the connected regions. Eventually the pipeline target detection algorithm was used to process the sequential images to detect the real small target automatically.

In this paper a residential area texture description based on the 3 x 3 region grey deviations is designed and the Gauss blur is applied to make the residential area in the texture character image possess accordant grey value and limited contrast relative to the background area so as to obtain self-adaptive threshold for image segmentation. And a skeleton processing is proposed to eliminate the road from the residential area. The experiment results of the semi-automatic extraction of the residential area in the remote sensing image with 3 meters ground resolution show this technique is very simple and effective to the semi-automatic extraction of the residential area and can meet the precision requirement of the mapping and surveying with satellite images.

Automatic extracting and updating road networks is a key work for updating geo-spatial information especially in developing countries. Aiming at the deficiency of the perceptual grouping based on geometric relation and based on the similar relation, a new perceptual grouping method that is so-called the perceptual grouping based on teh whole relationship is presented. In this method, all kinds of information including geometric properties, image attributes and other information are group based on the similarity rules. Based on this new grouping framework, the principles and the procedures of automatic road segments grouping are described in detail from two aspects: one is automatic perceptual grouping for similar road segments and another is automatic extended road segments for no-similar road segments. At last some discussion for new perceptual grouping strategy is given.

On the basis of analysis for degradation model of image due to uniform Liner motion, this paper deduces restoration model of image blurred by uniform Liner motion, and constitutes a kind of cost function of restoration. By analyzing the rule of cost function, the paper proposes an algorithm estimating motion parameter. The results show that the algorithm is very efficient to the restoration of image blurred by uniform. Liner motion when velocity on uniform Liner motion isn't very large.

This paper presents an event based soccer video retrieval method, where the scoring even is detected based on Bayesian network from six kinds of cue information including gate, face, audio, texture, caption and text. The topology within the Bayesian network is predefined by hand according to the domain knowledge and the probability distributions are learned in the case of the known structure and full observability. The resulting event probability from the Bayesian network is used as the feature vector to perform the video retrieval. Experiments show that the true and false detection rations for the scoring event are about 90% and 16.67% respectively, and that the video retrieval result based on event is superior to that based on low-level features in the human visual perception.

A new road recognition algorithm based on local statistical features and principal component analysis is introduced to improve whose robustness and adaptiveness. The weights of the principal component neural networks is trained with the aid of the algorithm of generalized Hebbian learning rule, and the input vectors of the local spatial features and image pixels value are transformed into feature vectors which are once clustered by K-means classifier, the road surface and un-road surface can be distinguished by the reference area finally. The simulation results confirm the fine robustness and adaptiveness of the newly proposed algorithm, especially, the improved performance to recognize road images affected by illumination variations or shadows.

In this paper, trademark is regarded as a combination of several geometric regions which have sharp edges. To retrieve such a combination, a method based on its member regions' shape and spatial features is proposed. Since the way considers the shape feature adn spatial relationship at the same time, so it can ensure the consistency in both the local and whole sides. Compared with the method of only using shape feature to retrieve the trademark, the results of experimentshow this way has higher precision and the output accords with people's visual feeling better.

Texture Mapping plays a very important role in Computer Graphics. Texture Synthesis is one of the main methods to obtain textures, it makes use of sample textures to generate new textures. Texture Transfer is based on Texture Synthesis, it renders objects with textures taken from different objects. Currently, most of Texture Synthesis and Transfer methods use a single sample texture. A method for Texture Synthesis adn Transfer from multi samples was presented. For texture synthesis, the L-shaped neighborhood seaching approach was used. Users specify the proportion of each sample, the number of seed points, and these seed points are scattered randomly according to their samples in horizontal and vertical direction synchronously to synthesize textures. The synthesized textures are very good. For texture transfer, the luminance of the target image and the sample textures are analyzed. This procedure is from coarse to fine, and can produce a visually pleasing result.

This paper combines the template-based method and color-based scheme to construct an adaptive skin-color model for human face detection in news videos. A heuristic rule-based decision tree is then employed to verify the resulting skin-color regions. The skin-color model comes from the sample pixels from the target video shots, so well tuned to adapt to various videos. It is a general scheme for color-segmentation, not depend on any pre-defined skin-color range. Our experiments shows that the face detection performance has been improved greatly, compared with the pure template based face detector. The face retrieval module is based on the Self-Eigenface method, where the Self-Eigenface space is constructed from the pseudo frontal faces obtained by region tracking.

In order to improve photolithographic resolution and focal depth of a projection photolithographic imaging system with an enough large numerical aperture, the basic principle of pupil phase-shift filtering has been investigated in detail. The mathematical model of filtering, the simulation and the photolithographic experiments have been carried. The theoretical analysis and experimental results show that both the photolithographic resolution and focal depth have been obviously improved with pupil phase-shift filtering. At the same time, the photolithographic window has been increased. Compared with amplitude filtering, the utilization of light energy and the potential ability to improve image quality of the phase filtering technique is fully excavated. It is an effective wavefront engineering technique for improving both photolithographic resolution and focal depth.

Waveshrink has been proven to be a powerful tool for the problem of signal extraction from noisy data. A key step of the procedure is the selection of the threshold parameter. Donoho and Johnstone propose of the threshold based on a SURE procedure for real signals. In this paper, we discuss the issue of threshold selection for complex signals in Waveshrink. We first review the threshold selection procedure based minimax thresholds and then propose to extend the use of SURE procedure for denoising complex signals with complex wavelet transforms. At last, an example is used to show that the extended SURE procedure is an effective method for denoising complex signals.

A new method for feature selection using radial basis function neural networks based on fuzzy set theoretic measure is proposed. The network's input values are all the membership of f feature values in a certain sample appertaining to C lcass (f: the amount of features, C: the amount of classes). Here, the fuzzy set theoretic π measure based on the normal distribution is used for computing the membership. Hence, there are f x C π measurements that are used as the inputs of the neural network. A radial basis function neutral network with increasing hidden nodes is trained for classification, which is believed to be able to perfectly simulate the nonlinear relevance among the inputs. And then, we set zero to the C input nodes concerning one feature (we call this input vector the revised input vector), which means as far as this feature is concerned, it belongs to none of the classes, which is considered to be the real delinking. The deviation between the output corresponding to the revised input vector and the expected output corresponding to the unrevised one is thought to denote the impact and the importance of this feature. Through this way, we may rank the features and select a suitable feature subset. Effectiveness of this algorithm is demonstrated on several sets of data, and compared with the effect of the feature-bsed node-prune MLP neural network method.

A novel face detection algorithm under hypothesis-verification scheme is presented, which includes three stags: skin region extraction, face candidate generation, and face candidate verification. This algorithm has several advantages: first, the skin chroma chart is fuzzily enhanced, which guarantees better discriminant power; second, through post-processing the extracted skin regions, overlapped regions are separated, which reduces the face detection complexity; third and the most important, maximum valley peaks from morphological operations are used as the invariant facial features for face hypothesis, which are more stable and accuracy than the commonly used valley block centers; fourth, to speed up the system, a multi-threshold fusion based image segmentation algorithm is proposed to constrain unreasonable candidates. At last, support vector machine is used for face verification, which is perfect for this task. For more than 1000 face images with different sizes, poses, expressions and lighting conditions, also including some gray images, the false rejection rate (FRR) is below 0.8%, false acceptance rate (FAR) is below 2.5%, and the average detection time is 2.55s. Experiments also show that the eye locations are very accuracy.

Based on analysis of many image change detection methods, a new method of multidimensional change template analysis is proposed in this paper. The new method combines the use of GIS knowledge and the advantages of other change detection methods. It has been used to detect change and update image database for the different temporal digital orthophoto maps (DOM). Test results show that this method is effective.

This paper presents an automatic video based face verification and recognition system by Support Vector Machines (SVMs). Faces as training samples are automatically extracted from input video sequences in real-time by LUT-based Adaboost and are normalized both in geometry and in gray level distribution after facial landmark localization via Simple Direct Appearance Model (SDAM). Two different strategies for multi-class face verification and recognition problems with SVMs, "one-vs-all" and "one-vs-another", are discussed and compared in details. Experiment results over 100 clients are reported to demonstrate the effectiveness of SVM on video sequences.

This paper presents a new method for locating eye accurately. In the first stage, the proposed method finds the coarse eye region from image using genetic algorithm based on the edge information and intensity distribution information. In the second stage, ellipse detection is employed to extract the boundary of the iris. The experimental results have shown that the proposed method can locate eyes accurately from the input image with complex background.

In this paper a new algorithm to locate the Elastic Labeled Graph is proposed for the face recognition approach based on Gabor wavelet jets. We extend Direct Appearance Model (DAM) to a hierarchical organization, which performs faster and more remote compared with the traditional graph localization method used in Elastic Bunch Graph Matching. A tracking recognition scheme is further discussed through employing the hierarchical DAM in a video sequence. Experimental results demonstrate the effectiveness of the method in locating and tracking the elastic graph.

This paper present an approach based on improved hybrid elastic model for deformable image registration. This method utilizes the linear spring net model for correspondence, and the thin-plate spline for non-rigid mapping. Compared with the original model, this improved method uses not only the intensity information, but also the weighted gray level histogram and the image gradient for similarity measurement. A multi-resolution strategy is involved to decrease the computing complexity and to increase the precision. Some experiments are performed on both synthetic and segmented medical images.

Independent Components Analysis (ICA) is an effective approach of blind source separation and has been received much more attention because of its potential application in signal processing such as telecommunication and image processing. Feature extraction of images has been also focused as one of prominent applications of ICA. Nine Stroke Density (NSD) feature extraction method will provide sufficient information to the recognition engine. Several other feature extraction methods are discussed and compared to stroke density method in detail. ICA extracts the underlying statistically independent components from a mixture of the NSD feature vectors. These independent components are feed into the neural netowrk for the recognition purpose. The experiment results show that ICA performs well for feature extraction and this proposed method is more effective in recognizing handwriting character than merely using neural networks directly.

The elliptic curve cryptographic random sequences as watermark are embedded in wavelet transform domain of the cover image. This algorithm takes advantages of the multiresolution feature of wavelet transform and non-relevant feature of the cryptographic random signal. The cryptographic random sequences are generated by the elliptic curve group and Galois Field function selected. The experimental results demonstrate that the scheme proposed is security, invisible and robust against commonly image processing techniques.

This paper presents a novel approach for mobile robot localization using monocular vision. The proposed approach locates a robot relative to the target to which the robot moves. Two points are selected from the target as two feature points. Once the coordinates in an image of the two feature points are detected, the position and motion direction of the robot can be determined according to the detected coordinates. Unlike those reported geometry pose estimation or landmarks matching methods, this approach requires neither artificial landmarks nor an accurate map of indoor environment. It needs less computation and can simplify greatly the localization problem. The validity and flexibility of the proposed approach is demonstrated by experiments performed on real images. The results show that this new approach is not only simple and flexible but also has high localization precision.

In this paper we obtain higher pattern analysis through the fusion of uncorrelated discriminant vectors, correlated discriminant vectors and kernelized discriminant vectors by some fusion theory and technology which is carried out by some estimation method of multi-feature. Based on some different features such as linear and non-linear features, correlated and uncorrelated features, one estimation method of multi-feature is proposed to fuse these different vectors. Finally experiments on human face recognition are carried out and prove our methods to be available.

We construct kernel uncorrelated optimal discriminant vectors(KUODV) for non-linear feature extraction and discrimination. Employing the uncorrelated optimal discriminant vectors(UODV) and kernel method, we propose non-linear generalization of uncorrelated optimal discriminant vectors, and then enhance the performance of original UODV. Human face recognition experiments show the utility of our new method.

In the field of data mining, it is often encountered to perform cluster analysis on large data sets with mixed numeric and categorical values. However, most exciting clustering algorithms are only efficient for the numeric data rather than the mixed data set. For this purpose, this paper presents a novel clustering algorithm for these mixed data sets by modifying the common cost function, trace of the within cluster dispersion matrix. The genetic algorithm (GA) is used to optimize the new cost function to obtain valid clustering result. Experimental result illustrates that the GA-based new clustering algorithm is feasible for the large data sets with mixed numeric and categorical values.

A new general method of the automatic selection of guide star, which based on a new dynamic Visual Magnitude Threshold (VMT) hyper-plane and the Support Vector Machines (SVM), is introduced. The high dimensional nonlinear VMT plane can be easily obtained by using the SVM, then the guide star sets are generated by the SVM classifier. The experiment results demonstrate that the catalog obtained by the proposed algorithm has a lot of advantages including, fewer total numbers, smaller catalog size and better distribution uniformity.

In recent years, the Asian dust storm project was carried out. One of tasks was to study dust rising mechanism in dust source area. Surface temperature condition was regarded as one of the important factors for dust rise. In the study we retrieved surface temperature by using NOAA/AVHRR data. Basedon the published articles, traditionally, split window algorithm was use to deriving surface temperatures in the case of our study area mostly desert area, there was only three field observation data available in Talimu basin, at Dunhuang and Changwu. It was very difficult to validate the results. However, there were 52 county wearther observation stations in the area. The data might be used as import data in artificial neural network calculation. Most success examples of remote sensing data classification by using neural networks were in the condition of network training and classifying in the same types of data such as spatial data. For the use different data type collected by different techniques system such as satellite system and ground weather observation data to training, to find rule and to direct classification could be more impersonal which was one of the nature of artifical neural network method. In our case 52 weather temperature data were used from 52 observation stations where they were also the same positions for collecting AVHRR 1b data CH2, CH4, CH5 thermal data. Both groups of data were applied as fundamental import data in for artificial neural network calculation. Finally resultant rule was applied for classifying 15000 x 3 pixels in the whole area. The result was more reliable than that of split window not only because uncertainty caused by variations of topography but also it was very difficult to validate in field.

A novel method of optimizing feed-forward neural networks using cascaded genetic algorithm is proposed in this paper. It adopts a hybrid encoding method, which architectures and connection weights vector of neural networks are encoded into binary code and real-value code respectively. The proposed optimizing method includes two cascaded evolutionary procedures in which the first mainly plays the role of fast search in constrained area and the second extends global exploration ability. The proposed method has represented a particular compromise between exploitation and exploration of searching optimized neural networks and enhanced the global search ability while using less computation. The experimental results have shown its good performance.

With the development of non-contact measurement in the close range photogrammetry, the use of the slide projector becomes familiar and frequent. In order to take full advantage of the slide projector in the procedures of diverse photogrammetric measurements, the slide projector needs to be calibrated at first. Namely, the intrinsic paramters of the slide projector have to be calibrated in advance. In this paper a flexible technique is proposed to calibrate the slide projector easily. The technique only requires an ordinary slide projector, a camera and a planar grid. The algorithm with 2D direct linear transformation (2D-DLT) and collinear equations is used to calibrate the slide projector. The operation method in detail and the algorithm are addressed systematically and entirely. First, the restricting condition among 2D-DLT parameters is worked out using the correspondence of 2D-DLT and collinear equation. Then, the decomposition of initial values of the slide projector intrinsic and extrinsic parameters using 2D-DLT is deduced. Finally, the slide projector calibration parameters are worked out by the whole adjustment. The feasibility and the exactness of the slide projector calibration technique put forward in this paper are verified by the results of real data.

Determination of image exterior parameters is a key aspect for the realization of automatic texture mapping of buildings in the reconstruction of real 3D city models. This paper reports about an application of automatic aerial triangulation on a block with three video image sequences, one vertical image sequence to buildings' roofs and two oblique image sequences to buildings' walls. A new process procedure is developed in order to auto matching homologous points between images in oblique and vertical images. Two strategies are tested. One is treating three strips as independent blocks and executing strip block adjustment respectively, the other is creating a block with three strips, using the new image matching procedure to extract large number of tie points and executing block adjustment. The block adjustment results of these two strategies are also compared.

The 3D reconstruction of video sequences in this paper need not to know the parameters and locations of the camera. First step of our reconstruction is feature point matching and stereo correspondence by Singular Value Decomposition (SVD). After getting the redundant information between image sequences, we introduce stratification algorithm to get 3D model. We at first get 3D model in projective space, and then get that of Euclidean space by using constrain information.

Advances in echocardiographic systems and computer applications have made three dimensional reconstruction of anatomical structures possible which open a new and fascinating field of color Doppler flow image, but the spatial shape and quantitative evaluation of mitral eccentric regurgitation is often difficult in the clinical setting. In this paper, we present a method to complete the 3D reconstruction of the mitral eccentric regurgitation, mitral eccentric regurgitation information was first derived from color Doppler flow images and then the mitral eccentric regurgitation velocity values was mapped according to the color bar in the images. With the proper method of interpolation and rendering, the experiment result of 3D visualization of mitral eccentric regurgitation is satisfying. Measurements from a 3D reconstructed flow convergence region may be superior to measurements from 2D color Doppler recordings to calculate volume flow, 3D mitral eccentric regurgitation reconstruction is possible and opens new possibilities in flow quantification. Futher study of the method may be very helpful in the diagnosis of heart diseases.

Many methods have been brought forward for reconstruction of 3D objects. Niem proposed an automatic method which can be realized in a simple measurement environment. However it used a planar calibration pattern that was unable to retrieve the depth of the points. This makes it difficult for the object whose top is on a plane to be perfectly reconstructed. In this paper, we propose an interactive method to solve this problem.

This paper proposes a 3D reconstruction method based on the decomposition of matrix. The method uses the Singular Value Decomposition (SVD) of the fundamental matrix, which leads to a particularly simple form of the Kruppa equations optimized by conjugate gradient method. The derivation doesn't need the somewhat non-intuitive geometric concept of the absolute conic. After the projective depths are estimated, the non-singular 4x4 matrix is obtained to realize the Euclidean reconstruction. Experimental results demonstrate the effectiveness of the proposed method.

To resolve the problem of approximating scattered points with closed boundary curves constraint, several methods has been analyzed and a central-point based method has been given here. The key process is to construct inner curves connecting the central point of whole region and middle point of each boundary curve, and divide the whole region into n piece of rectangular patches, so we can construct a base surface with Coons patch. An example of pentagonal surface is given to illustrate this scheme.

Reconstruction of 3D objects and scenes based on 2D images is one of the most important topics in computer graphics and computer vision. Traditional visual hull reconstruction methods cannot correctly handle the situation wehre part of the object's concave surface is not reflected on the 2D silhouettes -- for example, the inner surface of a cup cannot be recovered frmo its photo images at any angles. In this paper, a new algorithm is presented to solve this problem, focusing on the passive reconstruction with local application of active methods as a complementation. The basic idea is to first obtain a 3D model from the octree hull reconstruction method, then apply some assistant facility to get the local information of the concave surface, and finally combine the two pieces of information together to obtain a more accurate surface model of the object.

A new vision location method based on radical basis function networks (RBFN) is presented in this paper. It fully utilizes the excellent ability of RBFN to approach the nonlinear mapping and have a good performance of high learning rate and adapting the different environment generalization. It sets up a non-linear relationship between the space sample points and the corresponding image information by learning, instead of traditional calibration method, and can be used for 3D measurement. In our lab, it was applied to 3D vision location system based on a multi-linear photoelectrical sensor system. Then experiment proves that it can quickly realize high-accuracy space location. This method could be supplied as a new one to solve the 3D space location.

In this paper, we analyzed the latest algorithms and technology of multi-band image fusion[6,7], and studied the current stereoscopic display technology [1,4]. While the studying the match of dual-channel, dual-band image, we put forward the idea of dual-channel, dual-band image stereoscopic color fusion. The characteristics fo low-light image and near-infrared image were studied. The probability of low-light image and near-infrared image as stereoscopic image-pair was analyzed, and finally realization of stereoscopic color fusion was introduced.

This paper presents a novel method to reconstruct the internal structures of a mechanical part based on 2D thickness measurement from an Infrared (IR) measurement system. Conventionally, the internal structures are measured by X-Ray imaging techniques but those methods suffer from large measurement errors (higher than 0.125 mm). Using an innovative fixture, this new method first registers the 2D thickness measurement data with a 3D CAD model or a 3D point cloud representing the external feature of the measured part, and then reconstructs the internal features based on the thickness information from IR system. Experimental results shows this new method provides significantly high accuracy compared with X-Ray imaging techniques.

A filter for speckle reduction in SAT image is proposed. On each level of wavelet decomposition, three images are used. One is the original image, and the others are obtained by rotating the original image by 45° and -45° respectively, and so 12 subbands are gotten. In the 12 subbands, four subbands, HL subband and LH subband corresponding to the original image, and two HL subbands corresponding to the second and third image respectively, are used for edge detection, and the LL, HL, LH, HH subbands of the original image are used for synthesis. By using each point's four wavelet coefficients in the four subbands for edge detection, the edge detection property of the point on the original image is captured, and then the edges are detected by setting a proper threshold. And so, the speckle can be reduced while the edges being preserved well by setting the wavelet coefficeints in the synthesis subbands corresponding to the points not on edges to zero but retain the wavelet coefficients in the synthesis subbands corresponding to the points on edges. For detection of some oscillating edges, the filter is improved by combining with the traditional threshold method. Simulations on synthetic images indicate that the new filter performs better than the traditional wavelet domain hard threshold or soft threshold method.

In this paper a practical surface reconstruction algorithm is proposed to efficiently process very large medical dataset in general PC. By considering the conflict between memory consumption and traversal speed, we restrict the traditional surface tracking in single layer and thus get a better trade-off between them. We also use a compression scheme to store the generated mesh, which decrease the memory requirement considerably. For efficient rendering, we employ a triangle strips generation algorithm to decode directly the com-pressed mesh into triangle strip. The experimental results tested on visible man fresh CT dataset show that the proposed algorithm is very efficient in both extracting and rendering phase.

A method is developed for the detection and segmentation of spot targets at sea surface. Firstly, the Sea-Sky-Division-Line (SSDL), close to the horizon, is detected by wavelet-transform to mark out the Target Recognition Region (TPR), which can reduce the target searching range. A Row average grayscale substraction (RAGS) operation is employed to correct the blur caused by the non-linearity distribution of the temperature field. To repress the clutter in the background and increase the SNR of the image, a morphology Top-Hat filter is utilized. Then, the image is opening by selecting a proper structuring element to acquire a few potential target points. Through searching the maximal intensity and determining a threshold, most of the false alarms can be eliminated and the doubtful targets can be segmented. When the SSDL is visible, the real point-target can be retained according to the TPR and the false target can be discarded. Under the conditions of invisibility of SSDL for it is outsdie of the image or it is obscure due to the weather, the segmented target is the real target. The experiment result shows that the method can effectively detect and segment infrared point target in complex sea background.

The paper researched the theory of concept lattice and the algorithms of association rule mining based on concept lattice, introduced the methods into remote sensing image mining, analyzed and discussed the spectrum characteristics mining, texture characteristics mining, shape characteristics mining and spatial distributing laws mining, analyzed the application of remote sensing image mining, such as the automation classification, intelligent retrieval of remote sensing image etc., finally, discussed some research directions.

In staring Infrared Focal Plane Array (IRFPA) imaging system, the spatial frequency of the scene is always more than half of the sampling frequency, thus aliasing formed. Usually, micro-scanning apparatus are used to reduce the aliased signal energy. When micro-scanning, mechanical scan mirror often used to steer the field-of-view (FOV) of the imaging system over a fraction of the pixel distance. Mechanical equipment is usually large, heavy, complex, expensive and especially less reliability. While, liquid crystal can be used to make a non-mechanical light beam steerer for micro-scan. In this paper, we analyzed the infrared spectrum characteristic and the electro-optical reaction of liquid crystal, and introduced the realization of the light beam steerer.

The modulus maxima of a signal's wavelet transform on different levels contain important information of the signal, which can be help to construct wavelet coefficients. A fast algorithm based on Hermite interpolation polynomial for reconstructing signal from its wavelet transform maxima is proposed in this paper. An implementation of this algorithm in medical image enhancement is also discussed. Numerical experiments have shown that compared with the Alternating Projection algorithm proposed by Mallat, this reconstruction algorithm is simpler, more efficient, and at the same time keeps high reconstruction Signal to Noise Ratio. When applied to the image contract enhancement, the computing time of this algorithm is much less compared with the one using Mallat's Alternative Projection, and the results are almost the same, so it is a practical fast reconstruction algorithm.

A set of new invariant moment descriptors - wavelet moment invariants, which combine wavelet multiresolution analysis and moments invariants target recognition method, are proposed in this paper and their invariance also have been proved. Wavelet moment invariants take both advantages of the wavelet inherent property of multiresolution analysis and moment invariants quality of invariant to translation, scaling changes and rotation. Furthermore, studies of the effect of using different wavelet functions and their orders are carried out. Experimental results show that wavelet moment invariants derived from the wavelet function having proper vanishing moments, symmetry and compact support have the best discrimination performance.

Isodyne, as an experimental measurement method, can analyze the inner stresses without destroying the specimen. It is unnecessary to use high-power laser as light source, and the equipments of the experiment are simple. Moreover, the fringe of the pattern is clear and undistorted. Previously, making use of phase shift, phase of Isodyne fringe can be determined through four patterns. In this paper, wavelet-transform is applied to the phase analysis of Isodyne fringe, and all the results are satisfied. With the method of wavelet transform, we can determine the phase of Isodyne fringe automatically and accurately with only a single pattern. Furthermore, all the processing can be programmed, so it is easy to realize automatic analysis.

Triangular Bezier surface is widely used for modelling complex object. However most of geometric modelling systems do not support it. It is necessary to convert it into tensor-product Bezier surface. In this paper, a new conversion algorithm is proposed to convert a triangular Bezier surface into an optimal trimmed tensor-product Bezier surface by using polynomial interpolation. Then the proposed algorithm is compared with previous algorithms in both computational cost and numerical accuracy. The results show that the proposed algorithm has both computational and storage advantages over previous algorithms. Its numerical accuracy is comparable with the previous ones for cases of the degree less than 6.

The embedded zero-tree wavelet algorithm (EZW) is widely adopted to compress wavelet coefficients of images with the property that the bits stream can be truncated and produced anywhere. The lower bit plane of the wavelet coefficents is verified to be less important than the higher bit plane. Therefore it can be truncated and not encoded. Based on experiments, a generalized function, which can provide a glancing guide for EZW encoder to intelligently decide the number of low bit plane to be truncated, is deduced in this paper. In the EZW decoder, a simple method is presented to compensate for the truncated wavelet coefficients, and finally it can surprisingly enhance the quality of reconstructed image and spend scarcely any additional cost at the same time.

In this paper, an efficient VLSI architecture for biorthogonal 9/7 wavelet transform by lifting scheme is presented. The proposed architecture has many advantages including, symmetrical forward and inverse wavelet transform as a result of adopting pipeline parallel technique, as well as area and power efficient because of the decrease in the amount of memory required together with the reduction in the number of read/write accesses on account of using embedded boundary data-extension technique. We have developed a behavioral Verilog HDL model of the proposed architecture, which simulation results match exactly that of the Matlab code simulations. The design has been synthesized into XILINX xcv50e-cs144-8, and the estimated frequency is 100MHz.

In this paper, we propose a technique that combined template matching and support vector machine for road identification from high-resolution aerial image. It is a model-driven approach that combines both the local and global criteria about the radiometry and geometry of linear structures interested. In this approach, the road center point is extracted by utilizing the general road model. Then the road center point is used as initial point for the template matching through which the road segment is obtained. The road characteristic is learned through the support vector machine that is based on the statistical learning theory. The support vector machine is a powerful learning method thatit can get high classification accuracy without too much training sample. These properties can be applied for extracting the road characteristics from few road samples. The support vector machine is used to extract the true road segment and remove the false road segment. The proposed approach has been experimented on high-resolution aerial image and its performance is satisfied.

A new fast fractal encoding algorithm for the processes of searching and matching is proposed in fractal image compression in this paper. The number of domain blocks searched to find the best match for each range block and corresponding encoding time are much reduced by elimination domain blocks not searching using the current minimum distortion and variance difference between the range block and domain block. The algorithm produces a completely identical fractal encoding to that of the conventional full search in reduced time.

We proposed a new line detection method in noisy images using Mexican hat wavelet filters. In our approach, we applied the wavelet transform in a multiresolution sense by forming the products of wavelet coefficients at the different scales to locate and identify lines at different scales. In addition, we also considered shifting line locations through multiple scales for robust line detection in the presence of noise. We found that our approach leads to an effective method to form the basis of a line detection approach.

Relaxation matching is one of the most relevant methods for image matching. The original relaxation matching technique using point patterns is sensitive to distortions such as missing and spurious points and random errors. In this paper, under the condition that the number of spurious points is known, we present an improved point pattern relaxation matching technique which can get better performance than the original one. Experimental results with simulated images are given.

Propose one simple and efficient multi-level thresholding method. Basic dynamic is used to assess the reliability of thresholds. All possible thresholds are detected and sorted by assessment value calculated in water flooding process. Basing on the sorted threshold sequence, when level number changes, thresholds need not be recalculated, and multiple results can be got efficiently. Experimental results are satisfactory.

A new multple description selection/separation method for transmission of video bit stream over wireless channel is proposed in this paper. The method inserts transition frames according to the relative motion between two neighboring frames, and then divided the video sequence into two descriptions with independent prediction loops. The experimental resulst show that the method can help the decorder more quickly recovered from single loss or burst loss compared with previous method, and provides more stable and better quality for the reconstruction of video sequence.

Trademarks' retrieval has obtained more and more attention in recent research on content-based management and utilization of image database system. To retrieve the images, the key is to get the shape features. In this paper, a new method for extracting shape features of trademarks is presented. Based on the theory of information, the method uses images' entropy and invariant moments to capture the shape and spatial information of images. The algorithm is easy and the experimental results show its invariability with respect to translation, scale and rotation of objects. What's more, it also have the noise invariance.

Color is a useful piece of information in computer vision especially for skin detection. In this paper, we propose a novel approach for skin segmentation and facial feature extraction. The proposed skin segmentation is a method for integrating the chrominance components of nonlinear YCrCb color model. The chrominance components of nonlinear YCrCb color space were modeled using a subgaussian probability density function, and then the face skin was segmented based on this function. In order to autheticate the face candidates region, firstly texture information in face candidate regions would be segmented using mean and variance of luminance information, and then eye would be located by the PCA edge direction information, and finally, the others features, such as nose and mouth, also were detected using the geometrical shape information. As all the above-mentioned techniques are simple and efficient, the proposed skin segmentation based on nonlinear color space method is invariably of lighting and pose. In our experiments, the proposed method has been successfully evaluated using two different test datasets. The detection accuracy is around 98%, the average run time ranged from 0.1-0.3 sec per frame.

A hybrid method for image interpolation is proposed. The method consists of three different approaches: Circular arc or B-spline interpolation, linear interpolatino and human visual sensitivity based on interpolation. The image can be divided into three regions: linear smooth region, sharp edge region and human visual insensitive region. The method uses local variance and mean value to find different regions adaptively. The linear interpolation is used for linear smooth region. The human visual sensitivity based interpolation is used for human visual insensitive region and the circular arc or B-spline interpolation is used for sharp edge region. Experiments show that proposed method produces results that are more visually realistic than standard function-fitting methods.

In this paper, an approach of edge detection based-on multifractal is proposed. We apply the 2D wavelet transform modulus maxima (WTMM) method to characterize pointwise Holder regularity and the multifractal spectrum, so edge information can be extracted directly from them. Experiment results demonstrate that multifractal based edge detection has strong flexibility and good detection effect.

China - Japan joint dust project was started in 2001 and carried out a synchronous test to monitor the process of dust storm by using satellite data, to observe dust on sites and to collect total suspend particle (TSP), wind direction, wind intensity by using instruments, Si was tested in laboratory by analysis of collected TSP. LST (Land surface temperature) was retrieved from AVHRR data (96 day/night track imagery data). Resample 1:100,000 scale land use/cover data was used as reference layer during retrieving process. The comparison result of LST and TSP showed that there were good correlations. The result suggested that LST reflected some ground physical feature changes during dust storms and could be used as indicators for predict dust storm.

In this paper, the direct-sequence spread-spectrum technique is introduced to spread watermark signal. Then by utilizing a just noticed difference threshold in spatial domain based on visual masking effect of human visual system, watermarks strength and the embedded locations can be controlled in imagae adaptively. It was shown with experiments that the digital watermark not only can guarantee the transparency, but also has good robustness to JPEG compression, median filter, sub-sampling, etc.

Automated extraction of hierarchical catchments of river networks are fundamental to the automation of flow-routing management in distributed hydrologic models and to the morphometric evaluation of river network structure. An algorithm is proposed for automated extraction of hierarchical catchments from a river network database based on classiciation of river and constrained Delaunay triangulation network in this paper. At first the river network will be ordered by Horton's classification. Then triangulation network of this ordered river network will be constructed. This triangles in the triangulation network can be classified into several types based on their properties. These different types of triangles play an important role in analysis and building hierarchical catchments of river network. The algorithm has been tested in a test dataset.

In this paper, we present a method in the context of pattern characterization. This method is based on the analysis of closed contours of planar objects. The input contour is, first, separated into its x and y coordinates to generate two 1D signals. Both signals are then progressively low-pass filtered with a Gaussian kernel by decreasing the filter bandwidth. The output signals X and Y are then scaled so that the reconstructed contour and the original one can intersect. By doing so, we generate the so called IPM (Intersection Points Map) function that yields interesting attributes for pattern characterisation. The experimental results obtained by applying this method to various contours show that the IPM function is strongly related to the input contour and is rotation and translation invariant. It is also invariant under scale chance for a large range of scales. According to the experimental results, this function appears to be computationally very simple and to provide well-adapted features in the context of pattern recognition.

A new algorithm for image partition using irregular region in fractal image compression is purposed in this paper, which greatly increases the compression ratios achieved over traditional block-based partition. Also, due to the large search space involved (transformations and match), Similar Extension algorithms which are described are used to construct the irregular region transformations, and results for Similar Extension algorithms are shown. The results show that the algorithm of irregular region achieves almost double the compression ratio of simple block-based system at a similar decompressed image quality.

IHS transform was one of typical method for remote sensing data fusion. In recent years, newly developed method that combines advantages of IHS and Wavelet algorithms makes image fusion. In this case after the Wavelet substitution based on pixels or features, and then transforms inversely with IHS in Munsell color space. In this paper we introduce a high frequency substitution method to improve spatial resolutions of imagery. The procedure of the method introduced as flowchart, in which the dot line area is our newly added method. The resolution was greatly improved comparing original image. In cooperating with the demand of on going Minjiang river, Si Chuan, China. A 15m resolution PAN band and 30m resolution 7 bands of ETM data were selected for the method testing, the steps of method test showing in flow chart of this paper. In the future the dots area was our newly developed wavelet high frequency substitute. Improved NDVI imagery raised the quality for monitoring land cover change factor in the project of Return Farmland Back to Forest or Grassland.

A new lossless compression based on neural network is given by establishing special mapping Y, integral function and neural network. A high efficiency image compression based on wavelet and neural network is obtained by embedding in a good wavelet coding system with the new lossless compression. A new image compression based on fractal and neural network is also constructed by embedding in an efficient fractal image compression with the new lossless data compression based on neural network. Experiments show that these compressions are useful and efficient.

The wavelet transform is a powerful tool that cuts up signal or functions into different frequency components, and then studies each component with a resolution matched to its scale. However, how to study these components? This paper addresses the construction of morphological wavelets by combining wavelet with mathematical morphology. First, the multidimensional multi-channel lifting scheme, a general framework of multidensional morphological wavelet construction is presented. Then one-dimensional and multi-dimensional multi-channel median morphological wavelets are constructed wtih median operator.

This paper proposes an approach based on using lifting scheme to construct integer wavelet transform whose purpose is to realize the lossless compression of images. Then researches on application of medical image, software simulation of corresponding algorithm and experiment result are presented in this paper. Experiment shows that this method could improve the compression ration and resolution.

One common task of image interpolation is to enhance the resolution of the image, which means to magnify the image without loss in its clarity. Traditional methods often assume that the original images are smooth enough so as to possess continues derivatives, which tend to blur the edges of the interpolated image. A novel fast image interpolation algorithm based on wavelet transform and multi-resolution analysis is proposed in this paper. It uses interpolation and extrapolation polynomial to estimate the higher resolution informatoin of the image and generate a new sub-band of wavelet transform coefficients to get processed image with shaper edges and preserved singularities.

In this paper, a wavelet-based fractal image coding algorithm is proposed. The conventional fractal image coding in spatial domain is extended to wavelet domain by taking advantage of the self-similarities among different wavelet subtrees through proper affine transformation. This method is based on the combination of the theory of multi-resolution analysis with iterated function systems by introducing some effective block-classification schemes. The original image is first transformed into wavelet domain in which fractal compression and arithmetic coding are performed. By classifying D blocks and R blocks set in this domain, the approach can significantly reduce the computation complexity and encoding time. Meanwhile, the hybrid image compression algorithm obtains much better coding performance in terms of PSNR with error modification. This is the main advantage of this method. A set of experiments and simulations show the potentials of using these classification techniques in wavelet domain for futher improvements.

In this paper, we set forth the principle of Cosine Backscatter Model. In the model, and a new algorithm that doesn't omit azimuth angle and can extract DEM in mountainous area was introduced. First, the Radar image is divided into several regions by edge information using Lapalce algorithm. In one region, the image gray level changes slowly. Second, in the same region, we could assume that slope changes slowly, azimuth angle and range angle are affected by their neighbor pixels, the image gray level of pixel is changed by its neighbor pixels, azimuth angle and range angle were assessed from a seed. From known point, we get azimuth angle and range angle respectively by derivative; balance the value through iterative computation by ratio data and Cosine Backscatter Model. In neighbor regions, we get seed of gradient angle by average gray level of two regions and give amend index. From this point, we can get other point gradient angle same as the second step. Then we extract DEM in all regions. By applying this model, the DEM of Zhangbei of Hebei province were assessed. Through checking against the topographic map, the DEM error is little.

An endoscopic image retrieval system based on color clustering method has been presented in this paper. To reduce the color sensitivity to noise and the color histogram dimension and to enhance the accuracy and speed of retrieval, HSV color space is employed in this system and each Hue, Saturation and Value (Intensity) component have been quantified into 6, 8, 4 levels, respectively, because endoscopic images generally contain only a few of dominant colors, such as red, yellow or purple and HSV color space is most approximate to human perception, so the whole HSV space is divided into 192 (6x8x4) subspaces and each subspace is clustered as an index. So a color histogram with 256-dimension is used as indexing and the histogram similarity measure is also given at the same time. The algorithm has been successfully adopted by our endoscopic image retrieval system and the experiment with a database about 1000 clinical endoscopic images has demonstrated its effectiveness and rationality.

In this paper some properties of Foley-Sammon optimal discriminant vector (FSODV), by contrast with uncorrelated optimal discriminant vector (UODV), are discussed. Firstly the Fisher ratio of every FSODV must be not less than that of corresponding UODV and consequently sole FSODV will be superior to corresponding UODV. Secondly the correlation between feature components extracted by FSODV is an important factor. If high correlation is available between most of the feature components, the classification performance of FSODV will be remarkably inferior to UODV. However, if most of the feature components are only little correlative to each other, FSODV is comparative to UODV in classification.

A dynamical multi-objective evolutionary algorithm (DMOEA) is proposed. It is the first study of the dynamical evolutionary algorithm (DEA) in multi-objective optimization process. All individuals called as particles in a population evolve through a new selection mechanism. We combine the selection mechanism in DEA and the elitists strategy in existing evolutionary multi-objective optimization algorithms in DMOEA. The performance of DMOEA has been analyzed in comparison with SPEA2. The experimental results show that DMOEA clearly outperforms SPEA2 for the whole benchmark set. Moreover, a better convergence is sometimes observed in DMOEA for some functions of the benchmark set. The numerical experiment results demonstrate that the proposed method can rapidly converge to the Pareto optimal front and spread widely along the front.

In this paper, we propose a robust approach for object tracking in infrared imagery. Our method mainly applies the image intensity histogram distribution and intensity projection distributions and computes a likelihood measure between the candidate and the model distributions by evaluating the Mean Shift Vector. In addition, Gabor filters are applied here to enhance the contrast of the object with the background, and then the scale of the track window can be selected according to the variable object size. Our method greatly improves the accuracy of object tracking and can update the model frame by frame, which means the object model does not necessarily depend on that of the first frame. The robustness of our method is supported by several different infrared imagery sequences.

When compared with speech technologies in speech processing, automatic language identification is a relatively new yet difficult problem. In this paper, a language identification algorithm is provided and some experiments are conducted using OGI multi-language telephone speech corpus (OGI-TS). Then experiments results are described. It is shown that GMM-UBM is another efficient method to language identification problems.

Virtual endoscopy is meaningful for medical diagnosis and surgery. In this paper, a system framework for virtual endoscopy is proposed including automatic centerline extraction and view-dependent level-of-detail rendering techniques. Combining Hessian Matrix with distance mapping, our path planning method can generate accurate skeleton for virtual navigation. Furthermore real tim rendering can be achieved with our new view-dependent subdivision algorithm. The experimental results show the efficiency of our methods.

This paper describes the problem of labeling and completing an imperfect line drawing. Almost all of techniques about shape from contour require a perfect line drawing as their input. Experience indicates, however, that lines are often missing. We exhaustively studied all possible configurations of the imperfect line drawing and found that it is closely related to L-junction. W and Y-junctions in the imperfect line drawings are the same as they are in a perfect line drawing, L-junctions, however, are different because they could be degenerated from W or Y ones. In this paper, it is shown that the number of L-junction in the imperfect line drawing is 36. The 36 L-junctions and their corresponding W or Y ones from which they are degenerated are illustrated. We show how to utilize the line labeling and junction to predict and complete missing line.

In this paper, we proposed a new object-based coding algorithm by using wavelet transform to instead of the image encoder algorithm by using FGS in MPEG-4. The new object-based coding algorithm combines motion estimation with object-based 3-D wavelet transform for video coding in order to fully utilize the redundancy in the time domain. The shape-adaptive algorithm based on modifying boundary extension method of lifting scheme. A sequence of VOPs are fed into the motion compensated lifting (MCLIFT) wavelet coder which first decomposes the VOPs temporarily through MCLIFT filter, and then decompresses the VOPs spatially by shape adaptive lifting wavelet transform (SA-TWT). We encode the video and represent the stream as multilayer bit stream. The integrated transport-decoder buffer ensure the video be continuously transmitted. Losing package can be recovered by using re-transmission.

The problem of calculating the discrete Fourier tranform (DFT) acquired in polar coordinate system has been given considerably attention in many fields such as antenna, image registration and tomography. This paper proposes an improved fast DPFT algorithm aiming at 2D real data. In this paper, a Conjugated-like property of the conjugated sequences' Chirp-Z transform (CZT) in symmetric frequency section is proved which saves half of the computational complexity in CZT. The algorithm is suitable for real-time applications by only 1D calcuations in which the most steps are 1D FFT. The experimental results show the applicability and good performance of this approach.

Interfacial Fracture Mechanics is the problem which attracted researchers all over the world as the strength of the composite structure are depended on the mechanical behavior of the interfacial fracture. In the paper, the stress field and stress intensity factor of a bi-material beam with crack is obtained by model with initial carrier fringes. In order to obtaining phase value of which is related to the stresses, the Fourier Transform method and frequency shift are applied. It is shown that, using this method, the research of interfacial crack can be more simple, efficient and highly accurate. This technique is very significance for the local three-dimensional effects in dynamic problems of interfacial fracture mechanics.

A new reconstruction algorithm of computer tomography (CT) from a few views based on a neural network of Gaussian Machine (GM) is presented. The problem of image reconstruction is formulated as optimization under the criterion of maximum entropy, and a GM is then constructed to solve the optimization problem using simulated annealing technique with hyperbolic temperature adjustment. We demonstrate both the Simultaneous Algebraic Reconstruction Technique (SART) reconstruction of this image and the GM reconstruction using the same measured input data. The effect of noise in the projection data, projection angles and sample intervals are addressed. The results of numerical simulation show that this technique using the projection data obtained from four views with the projection angles 45°apart has fairly high accuracy (the average relative error is 0.03%) and good stability against noise.

The purpose of this paper is to describe a hardware model that can recognize Jawi handwriting manuscript via chain code technique. The model is divided into three modules: character segmentation, image processing (image enhancement, noise reduction, character recognition) and character id searching (via chain code). In the next stage, this model will be implemented using VHSIC hardware description language (VHDL).

While carrying out Optical Chinese Character Recognition, distinguishing the font between printed and handwritten characters at the early phase is necessary, because there is so much difference between the methods on recognizing these two types of characters. In this paper, we proposed a good method on how to banish seals and its relative standards that can judge whether they should be banished. Meanwhile, an approach on clearing up scattered noise shivers after image segmentation is presented. Four sets of classifying features that show discrimination between printed and handwritten characters are well adopted. The proposed approach was applied to an automatic check processing system and tested on about 9031 checks. The recognition rate is more than 99.5%.

The objective of this paper is to develop a rapid, objective, and easy method for recognizing wheat varieties, which is important for breeding, milling and marketing. The method can be used in place of the existing procedures to remove subjectivity from wheat variety recognition. In contrast to previous work, most of which has focused on wheat morphological characteristics, the features utilized in this paper are based mainly on kernel color. Varietal classification is performed by using Support Vector Machines (SVMs) method. More than 96% correct recognition rates are achieved with bulk samples involving 16 varieties representing a wide range of wheat varieties, wheat class, and kernel types. The proportion of single wheat kernels correctly recognized ranges from 87% to 93%. The results were encouraging since the method proposed here can be easily conducted in routine inspection.

Nowadays in intelligent transportation systems (ITS), information gathering depends heavily on visual information. Image processing technologies (IPT) play a key role. After a brief introduction of ITS, IPT is illustrated from three aspects: image sensor, image processing methods and image processing system. Among many applications for image processing in ITS, the paper presents a roadside example, licence plate recognition (LPR). Attention is centered around two aspects of LPR: plate character isolation and plate character recognition. Lastly, the paper indicates the trends of image-processing technologies in ITS.

The anisotropic reflectance of vegetation canopy is mainly determined by its spectral and structural features, and can be described by Bidirectional Reflectance Distribution Function (BRDF). In this article, we select the winter wheat from the beginning of April to the beginning of May 2001 at Shunyi county, north of Beijing, as the research object, to study its BRDF changing rule with the changing time. In the process we compute the structural scattering index (SSI) by inverting the semiempirical linear kernel-driven BRDF model, and analyze its relation with the leaf area index (LAI) of winter wheat. The results show that there is a clear linear relationship between SSI and LAI of winter wheat. So SSI can well be used to reflect the seasonal BRDF changing rule of winter wheat.

This paper presents a novel digital image watermarking scheme that is invariant to rotation, scaling, and translation (RST). We embed watermark in the log-polar mappings of the Fourier magnitude spectrum of an original image, and use the phase information of the original image to rectify the watermark positions. The scheme avoids computing inverse log-polar mapping (ILPM) to preserve image quality and avoid exhaustive search to save computation time and reduce false detection rate.

Observation by limited field observational stations is the major conventional method of research on snow and relevant problem, such as flood caused by melting snow. However, observation on only several ground observatories cannot provide enough information for large scale region accurately and timely. The advanced technology of remote sensing and geographic information system (GIS) is an effective tool to extract snow information and monitor snow change quickly and dynamically. This paper discusses the method of snow mapping and establishing dynamic snow monitoring information system by using multisensor, multispectral and multitemporal remote sensing data (NOAA/AVHRR, satellite borne SAR and TM).The study results show that the use of multisensor data and technique of GIS, combined with relative contemporaneous field observational data, enables the snow monitoring more rapid and accurate.

This paper addresses two problems of a ship handling simulator. Firstly, 360 scene generation, especially 3D dynamic sea wave modeling, is described. Secondly, a multi-computer complementation of ship handling simulator. This paper also gives the experimental results of the proposed ship handling simulator.

With the rapid globalization of market and business, E-trading affects every manufacture enterprise. However, the security of network manufacturing products of transmission on Internet is very important. In this paper we discussed the protocol of fair exchange and platform for network manufacture products E-trading based on fair exchange protocol and digital watermarking techniques. The platform realized reliable and copyright protection.

One important step in map vectorization is separation of geograhical information in a cartographic map. Such a fundamental task is not only time-consuming but also laborious. To get an overall understanding of this task and explain why it's hard to achieve a good result, a summary about the characteristics of color map segmentation is presented first in this paper. Then a new algorithm, featuring practicability and adaptability to the variation in brightness and contrast of images, is proposed directly based on RGB color model. Meanwhile, structure features are also used to improve the separation result by connecting broken lines. Experiments on many real map images prove that the algorithm realizes automatic segmentation of color maps and produces a fine result which can be used in automatic data collection without much further processing.

This paper introduces an automatic real-time gender classification system. The system consists of mainly three modules, face detection, normalization and gender classification. The LUT-type weak classifier based on Adaboost learning method is proposed for training both face detector and gender classifier, and a Simple Direct Appearance Model (SDAM) based method is developed to detect the facial landmark points for face normalization. This results in an integrated system with rather good performance. Experiment results on both pictures from World Wide Web and real-time video clips are reported to demonstrate its effectiveness and robustness.

Face technologies which can be applied to access control and surveillance, are essential to intelligent vision-based human computer interaction. The research efforts in this field include face detecting, face recognition, face retrieval, etc. However, these tasks are challenging because of variability in view point, lighting, pose and expression of human faces. The ideal face representation should consider the variability so as to we can develop robust algorithms for our applications. Independent Component Analysis (ICA) as an unsupervised learning technique has been used to find such a representation and obtained good performances in some applications. In the first part of this paper, we depict the models of ICA and its extensions: Independent Subspace Analysis (ISA) and Topographic ICA (TICA).Then we summaraize the process in the applications of ICA and its extension in Face images. At last we propose a promising direction for future research.

Artifical Neural Networks (ANN) has many good qualities comparing with ordinary methods in Land Suitability Evaluation. Based on analysis of ordinary methods' limitations,s ome sticking points of BP model of ANN used in land evaluation are discussed in detail, such as network structure, learning algorithm, etc. The land evaluation of Qionghai city is used as a case study, we know that ANN always can give more reasonable evaluation results from test.

We describe the principles of building a moving vision platform (a Rig) that once calibrated can thereon self-adjust to changes in its internal configuration and maintain an Euclidean representation of the 3D world using only projective measurements. The calibration paradigm is termed "Omni-Rig". We assume that after calibration the cameras may change critical elements of their configuration, including internal parameters and rotations. Theoretically we show that knowing only the relative positions between a set of cameras is sufficient for Euclidean calibration even varying focal length and unknown rotations. No other information of the world is required.

This paper discusses stereo photogrammetry analytic principle of the binocular sequence images and deduces the formula of the movement parameters estimate model. An aberrance correction model and sensors 3D spatial relationship calibration method is proposed. On this foundation, the common principle, calculation model and implement preceding and methods of the binocular sequence images aided by GPS/INS navigation are summarized. A method that used for positioning and orientation by GPS/INS assisted by motion analysis is proposed. Based on case of rapid scatter when GPS is lost, this method used the constraint offered by relative position and attitude from motion analysis to improve precision of position and navigatoin and constrain the scatter process. The experiment results of the vehicle navigatoin in GPS blocking case show the high navigation precision wiht the technique of the binocular sequence images aided GPS/INS navigation.

An algorithm to retrieve structure from long image sequence captured by a hand-held camera is proposed. Firstly, the long image sequence is divided into several subsets. Each subset has common feature points. Secondly, Euclidean reconstruction is obtained by factorization with all of these points visible in each image of a certain subset. Then results coming from different subset are brought into a common coordinate frame by the similarity transformations. Finally, global optimization is applied to refine the data and produce a jointly optimal 3D structure. A significant merit of the algorithm is that it can deal with the long image sequences with occlusions. The algorithm has been tested on real images with satisfactory results.

Morphological pattern spectrum is a useful shape description tool for quantifying the geometric shape feature of both binary and gray images. In this paper, a general frame of pattern spectrum is developed for both continuous-scale and discrete-scale based on the efficient and reduced redundancy multiscale image representation. A discussion for the basic properties of generalized pattern spectrum is presented in the paper. Algorithms for shape recognition and shape classification using continuous and discrete-scale pattern of images are proposed in Euclidian space.

In this paper, we address the problem of semantically generating hierarchical and meaningful content for soccer video by mapping low-level features to high-level semantics. Our goal is to construct a hierarchical and compact content abstraction of soccer video that can serve as an effective index table, allowing users to browse through lots of soccer videos in a flexible and efficient way. And we generated three-layer semantic hierarchies of soccer video according to characteristics of soccer video through bridging the gap between features and semantics. Some experimental results are presented and discussed in the paper.

SScalable reduced dimension face object segmentation and tracking (SRDOST) based on wavelet is presented in this paper. SRDOST algorithm is taken advantage of the characteristic of wavelet coefficeints multiresolution in the same direction, which makes SRDOST be applied to detect and track the video object of a reduced dimension image with much lower complexity and more sufficient accuracy. The number of image data at the lowest frequency subband is about one of (2^level)² to that of the original image so that the detection complexity at the lowest frequency subband may reduce greatly. It is important that SRDOST may be a multiresolution object segmentation algorithm based on wavelet transform, which may bring a family of video object sequence (VOS) with different resolutions. So SRDOST is a low complexity and efficient object segmentation algorithm. The proposed algorithm is to be integrated with our video object based wavelet color video coding with motion compensation algorithm.

An automatic SAR and optical image registration approach based on linear features and neural network is proposed. First, fuzzy linear feature extraction algorithm is used and common straight line segments between SAR and optical images are kept for matching primitives. Then discrete relaxation method is adopted to get acceptable matched primitives of two images and the crossing points of these matched line segments are taken as control points of image registration. Lastly, neural network is employed to realize theimage transformation and resampling. The experimental results are given and show that the proposed image registration approach can resolve the registration of SAR and optical images including long and thin objects effectively.

Face alignment is very important in face recognition, modeling and synthesis. Many approaches have been developed for this purpose, such as ASM, AAM, DAM and TC-ASM. After a brief review of all those methods, it is pointed out that these approaches all require a manual initialization to the positions of the landmarks and are very sensitive to it, and despite of all those devoted works the outline of a human face remains a difficult task to be localized precisely. In this paper, a two-stage method to achieve frontal face alignment fully automatically is introduced. The first stage is landmarks' initialization called coarse face alignment. In this stage, after a face is detected by an Adaboost cascade face detector, we use Simple Direct Appearance Model (SDAM) to locate a few key points of human face from the texture according which all the initial landmarks are setup as the coarse alignment. The second stage is fine face alignment that uses a variant of AAM method in which shape variation is predicted from texture reconstruction error together with an embedded ASM refinement for the outline landmarks of the face to achieve the fine alignment. Experiments on a face database of 500 people show that this method is very effective for practical applications.

Optical Flow computing doesn't require the rigorous corresponding relationship among features of sequential images, so this approach is widely used in computer vision field including detection and dynamic analysis of moving objects. But it is rarely used in infrared images because of the high noise levels of images. This article proposes a moving object pre-detection algorithm based on supervised learning, image pair difference significance test and minimum cost Bayes rule. This algorithm can not only efficiently be applied in indicating moving objects in infrared image sequences, but also in optical flow computing and behavior analysis of the moving objects.

This paper presents a novel parallel hardware architecture for MPEG-4 zerotree encoder. Under the architecture, a parallel processing of multi bit-planes is fulfilled through a preprocess until and multi-encoding units. The preprocess unit consists of mainly a bit-not-and and a bit-or logic circuits. It ensures sufficiently that efficient encoding in each bit-plane is performed independently. Each encoding until uses a fast technique to assign symbols by taking advantage of MPEG-4 zerotree coding symbol alphabet, and to select valid data to output using a ZTR address buffer.

Texture synthesis has a variety of applications in image processing. New algorithms emerged endlessly. Inspired by brushwork, we presented a novel algorithm for synthesizing textures from an input sample. First, EMD algorithm was introduced and texture structure of sample was extracted. Second, we synthesized texture structure with Patch-based sampling and MRF. For the spaces among structures, we extracted their irregular forms with distance transform and filled in them. Our method is not only simple and good qualitative but also robust.

The Wavelet-Domain Projection Pursuit Learning Network (WDPPLN) is proposed for resolving the difficult task that restoring image, which is blurred by multisource degraded factors image. The new approach combines the advantages of both the projection pursuit and the wavelet shrinkage technique. By separately processing wavelet coefficients and scale coefficients, the WDPPLN resolves the problem of restoring image very well, when little or not a prior information about the degradation is available. The WDPPLN estimates the degraded factor, which blurred the image, using Projection Pursuit Learning Network (PPLN). Also, it suppresses the noise using the soft-threshold of the wavelet shrinkage technique. The new method is compared with the traditional methods and the PPLN method in visual effect and objective evaluation criterion. Experimental results show that it is an effective method for restoring multisource degraded image.

Object-based segmentation of image sequences is one of the issues often arise in the world of video processing and communications. In this paper, a robust semiautomatic video object segmentation scheme is proposed. To facilitate users defining the initial object contour efficiently and accurately, an improved intelligent scissors is proposed by trading off the accuracy of original intelligent scissors and the simplicity of bounding box. To avoid the accumulated errors during object tracking, video sequence is firstly decomposed into video clips according to the rigidity of video object and the motion complexity. Then a snake-based bi-directional tracking is utilized to interpolate the video object planes (VOPs) of successive frames. Experimental results demonstrate that it can achieve better spatial accuracy and temporal coherency than COST211 AM, with about 10-22% improvement of spatial accuracy and almost the same temporal coherency.

A new algorithm for segmenting color colonscopic images by fusing color, brightness, spatial distance and texture information is presented in this paper. It makes the fractal dimension (FD) as the measurement for texture feature in images and applies a stochastic clustering algorithm that uses pairwise similarity of elements. The clustering algorithm that is based on a new graph theoretical algorithm for the sampling of cuts in graphs, can obtain the optimal number of clusters automatically. The complexity of our method is lower, and its stochastic nature makes it robust against noise. More than 40 colonscopic images have been used to demonstrate the effectiveness of this new algorithm.

Image segmentation is one of the most challenging problems in image processing. While significant progress has been made in gray-scale texture segmentation and color segmentation problem separately, the combined color and texture segmentation problem is less considered. In this paper, we use independent component analysis to extract local color and texture features for segmentation. Experiments compared with gray-scale texture analysis method show that the proposed method is more effective in segmenting complex color and texture images.

In this paper, an approach for the automatic extraction of linear feature, in particular roads, from digital aerial imagery is proposed. In some literature, knowledge based automatic road extraction were done with geo-referenced imagery which can automatically register old road map to new imagery as knowledge. Whereas the automatic geo-reference and road extraction are processed simultaneously in this approach, which can benefit and depend on each other. The implemented approach is based on Snake model and template matching with the new aerial imagery and old road map. The presented procedure does not need much manual interaction and therefore has the potential to be integrated into an automatic workflow. Potential applications of the approach are manifold, like automatic change detection of road and three-dimensional reconstruct of man-made objects such as road and building wtih some minor modifications.

This paper presents a texture segmentation approach based on Gauss-Markov random field(GMRF) model and multi-directional mosaics. Image texture is modeled by the second order GMRF model and the least error estimation is employed for the solution of model parameters. In order to improve the segmentation accuracy of uncertain area in boundary region between different textures, we introduced Laws energy masks and directional mosaics to obtain energy and orientation feature. And Euclidean distance approach is employed to classify different features. Experiments show that accuracy of texture segmentation can be improved.

Road extracted from satellite imagery have been used for many different purposes, e.g. military, map publishing, transportation, and car navigations, etc. Many method such as, neural network, Knowledge-based, Optimal search, Snake model, Semantic model, Road operator model, etc. was researched to identify road from satellite image, but because of complicated characteristics of road and image itself, and automated road network extraction still remains a challenge problem, and no existing software is able to perform the task reliably. This paper presents a hybrid method which combines Fuzzy-C-Means with back-propagation neural network and knowledge processing technique to detect roads in SPOT image. The basic idea of the paper is "easiest first" principal, and firstly focus to extract local salient road segments most easily and reliably, then use contextual knowledge and supervised back-propagation neural network model to extract fuzzy road segments among salient road segment, and then grouping these extracted pixel as seed point, candidate point, and not-road point, and then according to appropriate knowledge rule to traversal and join, guide the further road link in the whole image. At last, some post-processing steps are taken to refine the result. The resultant image shows this hybrid identification method performs better than only using knowledge-based method or neural network techniques.

The main characters of color topographic map are analyzed in this paper. A feedforward neural network is constructed and a heuristic learning algorithm is proposed to provide significant speedup and extraction of color from color topographic map. A new method of combining neural network with statistic techniques is given to make the algorithm of color map segmentation more effective and practical.

A new method of minimal fuzzy entropy segmentation is introduced. It adopts a new membership function for the consistency and concentricity in the object and its background. A new 2D fuzzy entropy thresholding method is also developed, which is based on 2D gray historgram. The gray values of every pixel and its neighboring region are used in this 2D method. The experimental results show that the minimal fuzzy entropy method is very useful in the segmentation of some images and the 2D method has a good performance of resisting noise and good robustness. The segmentatiaon of using 2D is much better than 1D for most images, and the new method can be easily extended to other 1D entropy imaging thresholding.

A novel unsupervised classification scheme called spatial fuzzy C-means clustering is proposed in this article. Based on conventional fuzzy C-means algorithm, our scheme takes spatial homogeneity into consideration by introducing spatial membership and applying SMNF, thus improved robustness against noises or outliers. Preliminary experimental results are also shown to demonstrate effectiveness of our method.

Image spectra calibration is of great importance for further processing and feature extraction. In this paper, an automated flat field reflectance calibration algorithm (AFFT) is proposed. This algorithm is an improvement to the traditional flat field transformation calibration. It is based on the fact that the so-called flat field is a flat block of high brightness and relative flat spectral response, and at a certain wavelength range (.e.g. 500-700nm) the brightness or radiance of the flat field is a certain multiple of the average spectrum of the image. Because the average image spectrum spectrum usuall is relatively flat, so a certain multiple of the average spectrum can be regarded as the criterion (or threshold) to select flat field pixels. So such parameters as wavelength range, multiple increment between flat field and the average image spectrum and number of the largest area block are set to determine the useful flat field so that an average spectrum of the flat field is obtained. By using this flat field spectrum as solar/atmospheric response, hyperspectral image can be calibrated to reflectance image. In the end, AFFT was validated by one PHI image acquired in Japan, 2000. It turns out that AFFT is effective to search all the flat fields which meet the fixed terms automatically and promptly, the spectra transformed by this method are much smoother and reliable to some extent.

These fusion methods such as IHS transform, Brovey transform and principal components transform could merged two optical image data of different resolutions - a high spatial resolution panchromatic image and a low spatial resolution multi-spectral image. But these fusion methods required the spectral range of the high spatial resolution panchromatic image equals or approximates to the spectral range covered with the multi-spectral image. This paper brings forward a new fusion method called FM that could merge two optical image data of different spectral range. This paper proposed its algorithm, firstly to filter on the panchromatic image, then to merge the remote sensing data applying algebra ratio. The fused production is more excellent at spectral preservation.

In this paper, a system for the segmentation and classification of the skewed document images with graph regions is proposed. In this system, the skewed angles of the document images are detected with a novel algorithm based on the morphological operation of Hit-or-Miss and the Hierarchical Hough transformation. To make the system valid for document images with graph regions, we proposed to introduce a middle point cut process to the traditional recursive X-Y cuts (RXYC) segmentation algorithm so that the graph regions can be approximated with a lot of small rectangles. The segmented regions are classified by two features of BWR and CC, which represent respectively the black to white pixel ratio and the cross-correlation between pixels of the sub-blocks. Experimental results have proved the fastness and the reliability of the system proposed in this paper.

In this paper a FLIR image segmentation algorithm based on genetic algorithm and fuzzy set theory was presented. Image processing has to deal with many ambigious situations. Fuzzy set theory is a useful mathematical tool for handling the ambiguity or uncertainty. A fuzzy entropy is a functional on fuzzy sets that becomes smaller when the sharpness of its argument fuzzy set is improved. The paper defined different member function for the object and background of the image to transform the image into fuzzy domain and chose Z-function and S-function as the membership functions for the object and background of the image respectively and threshold the image into the object and background by maximizing the fuzzy entropy. The procedure for finding combination of a, b and c is implemented by genetic algorithm with appropriate coding method to avoid useless chromosomes. The experiment results show that our proposed method gives better performance than other general methods with good real-time by using genetic algorithm.

With the development of modern science and techology, MEMS becomes an important branch. The micro operating system becomes an interested spot. During the micromanipulation process, observing the micro components by optical microscope is a crucial technology. Limited by the optical parameters, focus of the optical microscope is small. For example, focus of 10^x object lens is about 10 micron. The observation of some bigger micro objects, which size is tens or hundreds of times of 10 micron, are certainly impossibe by one time imaging. Some researchers had tried to extend the focus by improving the structure of microscopy, but the results are not satisfying. The optical microscope has CCD sensor as detector. Moving the object carrier, series adjusted focus images can be given. As the distance between CCD sensor and object lens is fixed, these images are with the same amplification ration. We can use computer to analysis these images for extending the focus. Through series images to extend focus, core of this method is analyzing and processing these series images, and at last composing one image, which is clear at each vertical depth. On this image, the position of each micro object can be measured easily. Sub-pixel processing technique makes the measure precision achieve micron degree. The distance, which is recorded while adjusting the object carrier, can help to locate the vertical position of micro object, experiments show that the locating precision could up to micron degree also. Using this method avoids changing optical system hardware, and is easy achieved. Clear image can be got in the adjustable range of object carrier. The range of focus is extended. The precision can be up to micron degree on 3D direction. So the method is useful on observing and measuring of micro object, has theoretical and practical value.

In this paper, a new fault diagnosis approach for the photovoltaic array, which is based on infrared and visible image fusion technique, is proposed. Firstly, the temperature difference and infrared characteristics are analyzed, and then the features of both infrared image and visible image are extracted. By comparing the features of infrared image with those of visible image, the abnormal operating regions covered with something are distinguished. A fuzzy fault diagnosis approach is introduced and implemented for other regions detected in infrared image but none in visible image. Experimental results show that the proposed approach is feasible and effective.

This paper discusses the existing three optimal band combination rules of hyperspectral remote sensing images. They are joint entropy, optimal index factor and Sheffield index respectively. Three bands of MODIS images data are combined arbitrarily according to the three rules, so the best three bands combination images of the three rules are acquired. On the basis of this, the three images are all classified in term of maximum likelihood classifier. Also, the influence of each band combination to the classification performance is discussed. The experiment result proves that the best classification performance of the MODIS images based on the three bands combination is the combination image based on optimal index factor.

Water is a kind of special natural background. The image acquired with laser imaging system for the underwater object always has some speckle noises caused by the backscattering of water and suspended particles, which gives birth to inconvenient to extract features of the image. In this paper, a set of laser underwater imaging system, which uses range-gated technique to avoid the backscatter and imaging distance up to above 20 meters, and its experimental results in boat-pool are introduced. According to the inherent mechanism of the underwater laser image, we propose a fractal-characters-based method for segmentation of the nature scene to find the artifact object from the image, which adopts region segmentation by Hausdroff dimension obtained by blanket covering method, and depends on the different distribution of the texture characteristic and multi-scale analysis to carry out the image segmentation. Experiments show the approach is suitable for texture segmentation and object finding in the image acquired b laser imaging laser system for the underwater object.

This paper mainly investigates the feasibility on the application of high-resolution AIRSAR in the topography of Surveying and Mapping, presenting the key procedures of data processing, extracting the basic mapping elements and evaluating the precision. Experiment results show that high-resolution and high quality AIRSAR image data could make its application extensive in the field of Surveying and Mapping.

SAR is a side-looking imaging mode and its image is very sensitive to the terrain shape. The little undulation of the terrain may induce the change of the image gray distribution and/or the texture characteristics. In this paper, the radarclinometry for extracting ditital elevation from single SAR image is investigated, which is based on the shape-from-shading principle developed in computer vision, consists in estimating the geometric parameters of a ground, from its radiometry and more precisely from the backscattered intensity coming from a piece of imaged ground. Firstly, the approaches for generation of the digital elevation from the SAR image data are discussed. Secondly the method of radarclinometry will be briefly described. The elevation reconstruction relies on the Lambertiam assumption for the terrain backscatter model. Then a single-line integral process is applied to calculate each pixel altitude, but it is still contaminated by noise. Finally the multi-line integral processing with various directions and the simulated annealing algorithm are respectibvely introduced to improve the single-line integral processing result. The presented experiment results promising in many geographic applications. This is an interesting technique of relief restoring, because it uses one single image only.

It has been demonstrated that Spotlight Mode Synthetic Aperture Radar (SAR) Imaging falls into a band-limited tomographic reconstruction problem in far field and small rotation angle situation. So far, several methods, such as two-dimensional inverse FFT (2D IFFT) algorithm and convolution back-projection (CBP) algorithm, have been developed to solve this problem. The purpose of our work was to investigate another method based on vector optimization (VO). In this article, mathematical model and algorithm are presented first, and then experimental results are shown to ascertain the superiority of this method over conventional ones.

A new algorithm for recognizing vehicle group targets in high-resolution SAR images has been given. This algorithm consists of two stages. At the pretreatment stage for images, the despeckling and segmentation for SAR images have been achieved by means of multi-look CFAR statistic detection and morphological processing for SAR images and the probable target chips are gained. The morphological multi-look filtering operators and target information resumption approach are developed and then the excellent image segmentation effect is implemented. At the target recognition stage, a group of individual and combinatorial invariable features, which can rightly express vehicle group targets in SAR images, are found. Individual vehicle targets are recognized by using individual features and the corresponding target classifying and recognizing approaches. Then the vehicle target arrays are found by using combinatorial invariable features and line detection approach. The method of parameter selection and the optimal parameters are also given. The recognition results for vehicle group targets have been shown.

Imaging spectrometers acquire images in a large number, narrow, contiguous spectral bands to enable the extraction of reflectance spectra at a pixel scale that can be used for identification. Many identification methods based on the spectra match technique have been developed. Such as spectral angle mapping, binary encoding. But these methods use all the data in the spectral dimension and compare the whole similarity between the reference and test spectrum. Sometimes two different kinds of spectrums may have big similarity, and this results in the wrong identification. There are also many algorithms using waveform characters for identification. However these methods maybe ineffective when the spectra have no diagnostic absorption feature. This paper introduces a new algorithm for identification based on diagnostic feature matching technique. Spectral matching technique and waveform characterization are combined for identification. Instead of matching test spectrum in all the wavelength range, this new algorithm emphasizes diagnostic features' location and only matches several diagnostic features in their most possible locations. To insure the idenfication of accuracy, spectral characters in terms of slope and asymmetry are used to check and verify. The algorithm is processed in four steps which will be described in the second part of this paper. In the third part, this algorithm is tested by identifying Alunite from AVIRIS image in Cuprite, Colorado. The result proved this new algorithm effective.

Using Terra/MODIS data of 2001, land cover classifications are studied in Kii peninsula based on pattern decomposition method (PDM). The pattern decomposition method is an analysis method developed in multi-dimensional reflectance space and is suitable for mixing analysis and sensor independent analysis. In this method, a digital number in multi-spectral satellite data is expressed by the combination of three standard spectral patterns multiplied by three decomposition coefficients. Use these three coefficients and vegetation index based on PDM (VIPD), land objects are classified and it is found that 76.8% of Kii peninsula is vegetation and is nearly same results as the values obtained using landsat/TM data of 1995. Our final goal of this study is to estimate net primary production (NPP) of vegetation in Kii peninsula using the results of land cover classification and to study annual change of land cover and NPP.

A new architecture for HIPAS (Hyperspectral Image Processing and Analysis System V2.0) was introduced in this paper which was modified and improved based on the first version of HIPAS V1.0. The comprehensive hyperspectral image analyzing system has been developed under VC++6.0 integrated development environment (IDE) and obtained perfect runtime efficiency and stability. The base architecture was specially designed and implemented to meet the requirements for the rapid preprocessing of imaging spectrometer data and easy prototyping of algorithms. Based on the modularized and object oriented software engineering construction, the architecture is compatible for other UNIX platforms with little modification. The most important components of HIPAS were presented in this paper including tools for input/output, preprocessing, data visualization, information extraction, conventional image analysis, advanced tools, and integrated interface to connect with general spectral databases. Some new methodologies for data analysis and processing were realized and applied to reach some valuable results based on the architecture including mineral identification, agriculture investigation, urban mapping etc. With an open storage architecture, HIPAS is entirely compatible with some advanced special commercial software such as ENVI and ERDAS and even the common image processing system Photoshop. At last, a strict and careful software test was carried out and the results were also analyzed and discussed.

In some complicated terrain area, such a loess plateau of China, it is very difficult to get higher accuracy of landuse classification only depending on the traditional spectral statistics methods, especially the image pixel size is much larger than the geomorphology units. In order to improve the image classification results, large scale relief map has been used to create the digital geomorphology model(DGM). DGM can be used to do the pixel unmixing works, specially reducing the influence of terrain shadow. Applying fuzzy mathematics theory, the DGM has been used to correct the digital image classification result, so as to create more accurate landuse map. In addition, this method is also helpful to find some minor objects in low spatial resolution images.

Projection Pursuit is applied to explore the potential structures and characters of the multi-dimension data through projecting the high dimensional data set into a low dimensional data space while retaining the information of interest. For the application of hyperspectral images analysis, the detection of small man-made objects is very difficult. But the man-made object can be viewed as anomalies in an unknown environment due to the fact that their spectral is different from those of the large known background. This paper presents a method to detect man-made objects of hyperspectral images based on Projection Pursuit. Also evolutional algorithm was developed in order to find the optimal projection index.

This paper presents a Prior Important Band (PIB) algorithm for the compression of hyper-spectral images. The PIB method endows some of the bands with high priority so that the quality of these bands after compression is better than other bands. The rationale behind this approach is that, the bands of a data cube have different amount of information. Some bands contain much more information and features than other bands. In the PIB algorithm, all bands are classified into four categories according to their importance and easiness for compression. For the simplicity of the compression algorithm, we choose spectral correlation and information amount as the main index. Bands of low spectral correlation and high information are selected as Important Bands. The benefit of this algorithm lies in that it treats the important bands with higher quality quantization, and other bands with comparatively low quality quantization, so that the information can be better preserved after compression. Experimental results illustrate that PIB hyper-spectral image compression algorithm would be suitable for most applications.

The number of plant disease spores in the air is a key factor for plant disease forecasting, so the realization of recognition and counting of spores by using computer has great significance. The digital image of plant disease is pre-procesed and the Fourier Description characteristics are studied for the first time in this paper. Based on this, reliable recognition of spores is realised according to 3σ rule. It can be seen from the experimental results that the probability of recognition for non-overlapping spores is 99.5 percent, and the probability of erroneous judgment that the noise and overlapping spores are recognized as spores is 0 percent.

In recent years, there has been a growing need for accurate, high fidelity scene simulations in the visible, infrared, microwave and other wavelengths. Based on a rigorous material classification and incorporating material attribute information, we generate wavelength independent texture maps for multi-spectral scene simulation. We calculate the sensor radiance value of every pixel, and change them into color or gray. If a single pixel in the texture contains more than one material, we mixture them based on their radiation attribution. According to area consistency and coherence across scan lines, an extended Seed Filling Algorithm is used in those areas with same or similar materials. These optical steps are performed repeatedly until a satisfactory classfication and mixture is found and the texture maps in a certain wave band are obtained. In this way we generate infrared textures from visible maps and different simulation scence textures at different time of day and under different environment conditions can also be obtained. Finally we give some examples of multi-spectral scene simulation, which are quite satisfied compared with the measured images.

A new hyperspectral image compression algorithm based on bit plane transform is proposed. The main idea of the bit plane transform is to decompose the hyperspectral image into a series of bi-level images which can be compressed more efficient. The gray code and band sequence optimization techniques are adopted to improve the compression performance in the preprocess stage. The main advantages of the algorithm are low complexity and easy realization.

An unsupervised classification method combining Principal Component Analysis (PCA) and Gaussian Mixture Model for hyperspectral image is proposed in this paper. It is based on the property that lower dimensional linear projections of high dimensional data sets have the tendency to be Gaussian, or a combination of Gaussian distributions as the dimension increases. The spectral dimensionality of the data is first reduced by a PCA linear projection; then the transformed data is modeled by a Gaussian mixture models, the parameters of the model are estimated using the Expectation-Maximimization (EM) algorithm in merge operations and the number of components is automatically selected based on Bayesian Information Criterion (BIC); finally the data after PCA transform is classified according to the mixture model. Applying the method to Push-broom Hyperspectral Imager (PHI) data shows that the method is quite effective without any a prior information.

It is an important factor for virtual endoscopy to extract center path. We introduce a new fast algorithm being suitable for single branch and multi-branch objects based on distance transform. A section plane including the next center point is found out according to the location relationship among the present center point on center path, object's start point and end point. The distance from inner voxels to boundary surface is computed in section plane, not in 3D space. The voxel point with the local maximum distance value is taken as the next center point, whose distance value in section plane is larger than its neighbors voxels' distance value. The method removing redundant branches on the center path of multi-branch object is also presented to ensure that the center path of single branch object or multi-branch object from start point to end point is unique based on algorithm. Our algorithm is faster and more efficient than other algorithm based on the distance transform.

Automatic analysis based on image content is a hotspot with bright future of medical image diagnosis technology research. Analysis of the fracture of skull can help doctors diagnose. In this paper, a new approach is proposed to automatically detect the fracture of skull based on CT image content. First region growing method, whose seeds and growing rules are chosen by k-means clustering dynamically, is applied for image automatic segmentation. The segmented region boundary is found by boundary tracing. Then the shape of the boundary is analyzed, and the circularity measure is taken as description parameter. At last the rules for computer automatic diagnosis of the fracture of the skull are reasoned by entropy function. This method is used to analyze the images from the third ventricles below layer to cerebral cortex top layer. Experimental result shows that the recognition rate is 100% for the 100 images, which are chosen from medical image database randomly and are not included in the training examples. This method integrates color and shape feature, and isn't affected by image size and position. This research achieves high recognition rate and sets a basis for automatic analysis of brain image.

The quantitative analysis software system for Myocardial Contrast Echocardiography (MCE) conforms to the Digital Imaging and Communications in Medicine (DICOM) standard and can be integrated into the Picture Achiving and Communication System (PACS). The MCE software system can measure the signal intensity of 2-D grayscale images and power Doppler images, draw the time-intensity curves to indicate the variation of the intensity of microbubbles scatting in Subendocardial layer and Subepicardial layer with the varying of myocardial segments, and estimate the hemodynamic parameters by nonlinear regression analysis. So the software brought the quantitative analysis of MCE to success.

Three-dimensional volume reconstruction has gained great popularity as a powerful technique for the visualization of volume datasets such as those obtained from X-ray, computed tomography, and magnetic resonance imaging in recent years. Local features play important part in the classification process for a variety of medical image analysis, computer-aided diagnosis, and three-dimensional reconstruction and visualization applications. By using high-order local statistic features detected by local block based moments, such as flat, round, elongated shapes, together with the local spectral histogram of textures, to act as classification criteria, a three-dimensional medical reconstruction method is proposed in this paper. A volume splatting algorithm by using the proposed classification method is implemented and relatively high-quality rendering results can be obtained when the proposed method is applied in medical reconstructions.

In this study, five classifiers, namely quadratic discriminant analysis, linear discriminant analysis, regularlized discriminant analysis, leave-one-out covariance matrix estimate and Killback-Leibler information measure based method are considered for classification of stellar spectra data. Because stellar spectra data sets are severly ill-posed, we first adopt some feature selection method such as principal component analysis to reduce data dimensionality. The input of the classifiers are those selected features, and the cross-validation technique is used to optimize the regularization parameters. Experimental results show that in most cases, regularized classifiers are high classification rates than that of quadratic discriminant analysis, but parameter optimization is time consuming. From experiments of exhaustive searching regularization parameter, it is found that in some cases cross-validation method is not always good in the selection of models.

Image watermarking has become a popular technique for authentication and copyright protection with the development of Internet and computer. However, current image watermarking approaches especially blind techniques are not strongly robust with respect to attacks or combinations of several attacks. In this paper a new intelligent second generation blind image watermarking technique is proposed, which adopts independent component analysis (ICA) for watermarking process. The characteristics of the human visual system (HVS) are incorporated into the watermark embedding, so that the watermark can be adaptive to the protected image. The edge of original image which extracted by Sobel operator is used as watermark in this paper. The watermark is rearranged by chaotic before watermark embedding in order to enhance the robustness of watermarking and the embedding process can be performed in any image domain, including spatial and transform domain. Watermark can be extracted correctly not merely be detected without any information about the original image and original watermark, and the accuracy of watermark extraction depends on the statistical independence between the original image, original watermark and the key. This proposed intelligent system can also extract multiple watermarks embedded in the test image one by one. Experimental results demonstrate that the proposed intelligent second generation watermarking technique based on ICA is robust with respect to attacks produced by popular watermark test software - Stirmark, including rotation, scaling, translation, skew, cropping, filtering, image compression, and combined attacks.

This paper proposes Variable Amplitude Fourier Series (VAFS). For a time-varying signal, the amplitude of VAFS vary as function of the time, which can be used as the tool of time-frequency localization. Based upon VAFS, Local Approximate Index(LAI) is proposed. LAI can be used to measure local maximal freuqency of signal. By using LAI, signal of time-varying can be segmented by frequently and time-invariant interval (TII) of signal can be obtained. Re-sampling of signal and speech segmentation can be effectively processed by using VAFS and LAI.

In this article we propose a hierarchical partition of the feature space based on statistical information on each dimension, and then use mean-shift to properly fuse the obtained super-cubics to reveal the genuine data structure. It not only greatly reduces calculation, but also provides a desirable priori knowledge for bandwidth selection.

An algorithm for watermarking of vector map based on discrete wavelet transform is presented in this paper. The watermarking information is represented by a bi-valued sequence. We embed the watermark repeatedly in the magnitude of the DWT coefficients vertices, which are extracted from the map. Considering the precision of the watermarked map might be violated and the map's usability might be damaged, the intensity of the watermark should be adjusted in the spatial domain until it meets the precision standard. Experiment results show that the watermark is significant but not violate the map's precision. The proposed algorithm is robust against some attacks.

Scene matching navigation based on multisensor image fusion is studied in this paper. Pixel and low feature -level fusion images of optical and IR images are used as real time images matching with optical satellite images as base images, in which linear superposition, nonlinear operators and multiresolution image fusion approaches are adopted to acquire the fused gray and edge strength images. Gray and low feature -level scene matching schemes are also employed to execute the scene matcing simulation experiments on the real flight image data, in which scene matching methods are general CCF and MAD algorithms. The experimental results are given to compare the matching performance when taking different fusion images as real time images under certain matching schemes. The scene matching results based on single sensor images are also given for comparison with the results based on multisensor fusion images.

In this paper, we first discussed the wireless multimedia transmission system and the relations among source distortion (D_s), channel distortion (D_c), and the end-to-end distortion (D). Then we focused on computation of D_c and based on a simple error concealment scheme proposed an effective channel distortion estimation algorithm. Simulation results have shown the high resolution of this estimation algorithm, which will be important for unequal error protection and joint source-channel coding.

Speckle suppression is an important step in syntetic aperture radar (SAR) image processing. This paper addresses the problem of reducing speckle in SAR images by employing stack filters. Median filters have been used to successfully reduce non-Gaussian noises and impulse noises. Stack filters are a very large class of nonlinear filters that possess the threshold decomposition and the stack property. In this paper, stacking median filters are introduced. An algorithm for SAR speckle suppression based on stacking median filters is proposed. Threshold selection is also discussed. Experiments are performed using SAR images at different threshold differences to show the proposed algorithm's effectiveness.

This paper presents a new filtering algorithm using multi-threshold decomposition filter (MTDF) to solve the contradiction between impulse noise attenuation and image detail-preserving. The algorithm firstly uses Otsu’s threshold segmentation (OTS) to segment the image into four parts by steps. Then the matrices of the impulse noise location can be obtained respectively using the threshold decomposition multistage median filter (TDMMF). At last, median filters separately according to the matrices filter the noise pixels in the original image. The results indicate that the method has not only the good ability of de-noising but also better properties in preserving details than median filtering.

Load-stepped method is a new full field automatic photoelasticity image processing method which can obtain the phase value directly. The principle of the load-stepped photoelasticity technique is introduced. The computer simulated photoelasticity images are used to describe the method and the results are satisfied. The authors expect the method will play an important role in dynamic photoelasticity image processing.

This paper proposes a new method to denoise images corrupted by Poisson noise. Poisson noise is signal-dependent, and consequently, separating signals from noise is a very difficult task. In most current Poisson noise reduction algorithms, noise signal are pre-processed to approximate Gaussian noise, and then denoised by a conventional Gaussian denoising algorithm. In this paper, we propose to use adaptive basis functions derived from the data using modified ICA (Independent Component Analysis), and a maximum likelihood shrinkage algorithm based on the property of Poisson noise. This modified ICA method is based on a denoising method called "Sparse Code Shrinkage (SCS)" and wavelet-domain denoising. In denoising procedure of ICA-domain, the shrinkage function is determined by the property of Poisson noise that adapts to the intensity of signal. The performance of the proposed algorithm is validated with simulated data experiments, and the results demonstrate that the algorithm greatly improves the denoising performance in images contaminated by Poisson noise.

The phase noise filtering of interferogram is one of important procedures in INSAR data processing. Any improper altering of the wrapped phase may influence the quality of derived DEM, because the interferometric phase contains the topographic information. A new adaptive method combined with Media filter and gradient-based adaptive filter is presented in this paper. The filtering action is conducted in the real and imaginary component of the interferogram respectively before the phase image is formed by means of the arc tangent operator.

In this paper, we present a new method by using 2-D discrete multiwavelet transform in image denoising. The developments in wavelet theory have given rise to the wavelet thresholding method, for extracting a signal from noisy data. The method of signal denoising via wavelet thresholding was popularized. Multiwavelets have recently been introduced and they offer simultaneous orthogonality, symmetry and short support. This property makes multiwavelets more suitable for various image processing applications, especially denoising. It is based on thresholding of multiwavelet coefficients arising from the standard scalar orthogonal wavelet transform. It takes into account the covariance structure of the transform. Denoising is images via thresholding of the multiwavelet coefficients result from preprocessing and the discrete multiwavelet transform can be carried out by threating the output in this paper. The form of the threshold is carefully formulated and is the key to the excellent results obtained in the extensive numerical simulations of image denoising. The performances of multiwavelets are compared with those of scalar wavelets. Simulations reveal that multiwavelet based image denoising schemes outperform wavelet based method both subjectively and objectively.

With urban and township development and E-Government program promotion in China city remote sensing as base data has developed rapidly. The technique demands in accuracy and effective edge detection and extraction from higher resolution image become important focal area. In the current popular image processing software packages there are some existing edge detection convolution kernels suchc as Sobel, Robert, Prewitt, Kirsch, Gauss-Laplace kernels. In general the kernels all work based on algorithm of convolution kernel in spatial territory of the image. However, satellite sensors capture spatial and spectral signatures of surfaces at same time. Use of both spatial and spectral features to establish a edge detection process is a new notion for achieving more accuracy results. In the paper we introduce a spatial and spectral integrated method which is designed in four stages. The result suggests that four stages process can achieve more cleanly and accuracy edges of city construction than that results of using other algorithms. The procedure is summarized in figure 1.

Edge detection is an indispensible step in the computer vision and object recognition, because the most fundamental characteristic for image recognition is the edges of an image. Traditional edge detection algorithms have had fairly limited application in ultrasound medical imaging. High levels of speckling present in ultrasound medical images make accurate edge detections difficult. In this paper, a novel edge detection method based on mathematical morphology is presented. First, according to the statistical characteristic of the speckle, the speckle filtering method based on multidirectional morphological structure is presented. The experiment shows that the speckle noise has been filtered and image details have been preserved by the method based on local statistical property of the image; second, after a series of morphological transforms, the edge detection result shows that this method is reliable and effective. Certainly the speckle filtering and the edge detection are not the final objective and would like to be proved and perfected in the futher research such as 3D reconstruction of echocardiography.

An improved live-wire algorithm is presented for interactive weak edge detection and extraction. In comparison with live-wire algorithm, the proposed one, with the same complexity as the original one, greatly improves its performance in edge detection of region of interest (ROI). Meanwhile, the modified algorithm overcomes the following drawbacks of the traditional one: (1) rather sensitive to noise, (2) inefficient to distinguish between the strong and the weak edges, (3) inapplicable to detect sharp edges. The experimental results illustrate that the proposed algorithm is really of good performance over the traditional one in three aspects.

Line detection is one of the most long-lasting problems in image processing. In this paper, we address the problem of detecting strong/weak lines and long/short lines in gray level images simultaneously. Our technique consists of three steps: image gradient enhancement, modified Hough Transform (HT) and line fitting. The gradient direction is fully exploited in our technique, which provides additional cues for weak line detection. Experiments on various images are presented to verify our approach. The results show that our technique has a superior detection rate than conventional HT algorithm, especially for short and weak lines.

Integral image as an intermediate image representation can be used to calculate the sum of gray level in rectangle quickly, based on which this paper presents a novel partial matching error function based on sub block mean. The optimal sub block division is a key to the matching error function based on sub block mean and is determined to be 4 under the hierarchical block matching. Experiments show that the matching error function based on sub block mean is superior to both the full matching error function and the matching error function based on sub sampling in terms of the motion estimation quality and speed. The matching error function based on sub block mean with the content sub block division guarantees the almost same wasting time to the different matching images, which is very suitable for the real-time applications such as video compression.

Based on Immune Genetic Algorithm (IGA), a novel algorithm for image segmentation is presented in this paper. Utilizing functions of self-adaptive, antigen recognition and memory of immune mechanism, this algorithm combines genetic operator to segment image. The results of experiments indicate that it is an effective new method for image segmentation.

A novel automatic Video Object (VO) segmentation method is presented in this paper, which is based on intersection of frame differences. Horizontal scan is used to acquire coarse VO mask and edge detection is performed on VO boundaries to remove uncovered background contained in the intersection. And morphological operator open is applied to smooth VO contours after extraction. Experimental results show that it is accurate and especially efficient, and can wonderfully meet the real-time requirements of applications such as stationary camera video surveillance.

Level set methods offer a powerful approach for the medical image segmentation since it can handle any of the cavities, concavities, convolution, splitting or merging. However, this method requires specifying initial curves and can only provide good results if these curves are placed near symmetrically with respect to the object boundary. Another well known segmentation technique - morphological watershed transform can segment unique boundaries from an image, but it is very sensitive to small variations of the image magnitude and consequently the number of generated regions is undesirably large and the segmented boundaries is not smooth enough. In this paper, a hybrid 3D medical image segmentation algorithm, which combines the watershed transform and level set techniques, is proposed. This hybrid algorithm resolves the weaknesses of each method. An initial partitioning of the image into primitive regions is produced by applying the watershed transform on the image gradient magnitude, then this segmentation results is treated as the initial localization of the desired contour, and used in the following level set method, which provides closed, smoothed and accurately localized contours or surfaces. Experimental results are also presented and discussed.

Matching features computed in images is an important process in multiview image analysis. When the motion between two images is large, the matching problem becomes very difficult. In this paper, we propose a contour matching algorithm based on geometric constraints. With the assumption that the contours are obtained from images taken from a moving camera with static scenes, we apply the epipolar constraint between two sets of contours and compute the corresponding points on the contours. From the initial epipolar constraints obtained from comer point matching, candidate contours are selected according to the epipolar geometry, the linear relation among tangent vectors of the contour. In order to reduce the possibility of false matches, the curvature of the contour of match points on a contour is also used as a selection method. The initial epipolar constraint is refined from the matched sets of contours. The algorithm can be applied to a pair or two pairs of images. All of the processes are fully automatic and successfully implemented and tested with various synthetic images.

Searching object's centroid fast and correctly is the important part in pattern recognition and object tracking. So it is crucial to search centroid fast and correctly. Song suggests a method of searching centroid with "City Block Distance" template and difference image. But the method of Song has two disadvantages, so an improved fast algorithm for searching centroid of object in binary image is proposed in this paper. Searching time and area are reduced greatly by the improved algorithm. It is turned out that the improved method is faster than not improved one by the experiment.

For the traditional method of hyper-plane segmentation, the location of hyper-plane in data space was given by statistical method. In the case of the statistical value of regions is smaller than in the region, the statistical method was not effective. The character of genetic algorithm is global searching optimally. Taken this mathematical advantage the location of Hyper-plane could be located easily. In this paper, EOS/MODIS imagery data is used to test this method. The result is proved that Genetic Algorithms-Hyper-plane is better than MLC method by using same training data.

This paper proposed a new stable algorithm based on feature point matching, which can be used in image registration under shift and rotation. The new method firstly extracts feature points in the first image of an image sequence and then searches the matching points in the consecutive images. Different from conventional methods, it adopts constrain of distance to revise matching and removing false matching, and also adopts projection of pixel to filter feature points and choose many possible matching points for one reference feature point, which makes it fast and reliable. Experiments prove this method is practical and reliable.

Recently, SAR interferometry (InSAR) is popular in low-resolution DEM generation and high precision deformation measurement. There are many uncertain errors in InSAR processing that cannot be modeled and eliminated, such as tropospheric delay, phase unwrapping error, baseline error, and propagating noise. As we cannot decrease these errors just by SAR data processing, many other means have been presented to solve the problems. GPS signal and SAR signal are microwave of the same kind, and have the same attribute in propagation, so we can integrate GPS data into InSAR data to eliminate the same propagating errors in the aerosphere. What is more, GPS can observe high precision deformation better than differential InSAR has done. It is valuable to compare the results derived from these two methods. Researaches undertaken by some institutes on integration of GPS and InSAR have shown a prospective future. Therefore, this paper aims to analyze the methods of GPS and InSAR integration and apply them into the case study of city subsidence.

At present, information fusion is widely used in object recognition and classfication, since this technique can efficiently improve the accuracy and the ability of fault tolerance. This paper presents a new method of information fusion system, which is used for offline Chinese signature verification. In the feature fusion process, different features are adopted to syncretize new features; in the report fusion process, the neural network is selected and used to adjust the weight. By comparing the experimental results of fault diagnoses based on fusion data with that on original separate data, it is shown that the former is more accurate than the latter.

A new stereo matching scheme based on image edge and genetic algorithm (GA) is presented to improve the conventional stereo matching method in this paper. In order to extract robust edge feature for stereo matching, infinite symmetric exponential filter (ISEF) is firstly applied to remove the noise of image, and nonlinear Laplace operator together with local variance of intensity are then used to detect edges. Apart from the detected edge, the polarity of edge pixels is also obtained. As an efficient search method, genetic algorithm is applied to find the best matching pair. For this purpose, some new ideas are developed for applying genetic algorithm to stereo matching. Experimental results show that the proposed methods are effective and can obtain good results.

In video sequences, object movement causes regions occlusion (to-be-covered or uncovered), which seriously decreases the accuracy of object segmentation. This paper presents a novel video object segmentation algorithm that can identify the occlusion regions. The background rebuilding technique is used to construct a reliable background image from the accumulated frame difference information. THe intial moving object segmentation is finished by the difference between the background image and the current frame, and the occlusion regions are detected. Then, the initial result is spatially segmented into homogeneous regions, and a method based on region labeling distinguishes between moving object and background regions. Experimental results for several MPEG-4 test sequences demonstrate the effectiveness of the proposed approach.

Feature extraction is an important step before object detection and pattern recognition are conducted. In this paper, we endeavor to find a new way of feature extraction. As well known, the fundamental of computerized tomography (CT) is image reconstruction from projections. And a famous image reconstruction algorithm is filtered back-projection. Since the filtered projections can reconstruct the original object, it can be inferred that they can also represent the object. One of the contribution of this paper is that the idea image reconstruction from projections is adopted. To reduce the pattern dimensionality without loss of the ability of characterizing the object, the fractal dimensions of all the filtered projections are computed. These fractal dimensions form a feature vector from which pattern recognition can be done easily. Preliminary results have demonstrated that the proposed approach is a promising method for feature extraction.

In this paper a high -speed target tracker is developed, which is composed of a frame buffer digital camera that operates at the frame rate of 200 frames per second, and a digital image processing board based on the high speed float digital signal processors, TMS320C6701. The design and realization of laboratory validation system, the simulation of target tracking, the image interface between camera and the TMS320C6701, and the algorithms flow, is given in detail.

A new method for designing Gabor filter banks with various frequencies and orientations is proposed to perform feature extraction. A feature decomposition modular neural network classifier based on the feature extraction by the Gabor filter banks is presented for object recognition problems and the influence of band limited noise.

To enhance the tampering localizability of image authentication, a new fragile watermarking scheme, which exploits the perturbation in reverse processing, is proposed. In verifying data integrity, the new method performs the reverse processing of watermark embedding. Typically, it de-filters the distributed version or equivalently solves an embedding equation. If any tampering happened, the output of the method perturbs because the manipulated data, which can be regarded as the observing error, is drastically enlarged by such processing. The perturbed values indicate the degree of tampering, and their positions directly draw the shapes of the manipulated areas. Compared with the mostly used block-based methods, the new method localizes the tampering almost pixel-wise instead of block-wise. It also supports the popular adaptive embeddin, which does not evenly embed the watermark.

In the target recognition system using Joint Transform Correlator (JTC), real-time displaying spectral images in electric addressed Liquid Crystal Display (EALCD) is required, under the work condition of non-affecting image capturing and processing, the above requirement can not be satisfied by the graphic device interface. With the aid of DirectDraw, image data can be captured, transmitted and converted. When the spectral images are displayed at a high speed, the high quality of image can still be produced.

Face recognition technology (FRT) has numerous commercial and law enforcement applications, especially in video surveillance. The primary task at hand, given still or video images, requires the identification of one or more persons using a database of stored face images. This paper first introduced a new face coding and recognition method based on integrated shape and texture features. And the dimensionalities of the shape and texture spaces are reduced using generalized KL transform. Experimental results show that this method does not sensitive to the pose and expression of human faces.

Detection of low observable small targets from a sequence of InfraRed (IR) images is an important research area for image processing applications, but the difficulties in detecting small targets arise from the fact that they are not easily discernable from the background clutter. In this paper, the author presents a detection method based on Wavelet and Higher-Order Cumulant Filter (WHOCF). Firstly, the proposed method suppresses the background clutter with wavelet filter, then uses a Cumulant-Based Adaptive Filter (CBAF) in the further processing of the wavelet sub-band, which can suppress noise and preserve the candidate targets information effectively, and improve the Signal to Noise Ratio (SNR) greatly. Finally, a sequence images processing method is used before detection to enhance the performance of the detector. The experimental results with real world image sequences illustrate the effectiveness of the proposed method.

A new concept that is the Discontinuous Frame Difference in image sequences is proposed in this paper, and is applied to the Optical Flow Algorithm. The modified Optical Flow Algorithm overcomes the shortcoming of the traditional Optical Flow Algorithm that cannot detect the small moving target whose moving displacement is less than one pixel between two continuous frames. The infinite norm of the Discontinuous Frame Difference Vector is used to preprocess the image sequences to get rid of most of pixels taht are not the pixels of the moving target. And then the instantaneous velocities of the pixels remained by preprocessing is calculated by the Optical Flow Algorithm. If the pixels in an area have the moving continuity and consistency, a moving object is determined. For the preprocessing is able to get rid of most of pixels, the calculation quantity of the Optical Flow is reduced greatly. But the preprocessing is probably to lose some candidate pixels of the moving target, so the Gray Intensity Analysis is used to find these pixels back again. The Discontinuous Frame Difference Optical Flow Field Algorithm can be composed of the parallel structure system, which can detect different kinds of moving objects with different velocities. The experiment result proves the effectiveness of the method.

Anchorperson shot detection is of significance for video shot semantic parsing and indexing clues extraction in content-based news video indexing and retrieval system. This paper presents a model-free anchorperson shot detection scheme based on the graph-theoretical clustering and fuzzy interference. First, a news video is segmented into video shots with any an effective video syntactic parsing algorithm. For each shot, one frame is extracted from the frame sequence as a representative key frame. Then the graph-theoretical clustering algorithm is performed on the key frames to identify the anchorperson frames. The anchorperson frames are further refined based on face detection and fuzzy interference with if-then rules. The proposed scheme achieves a precision of 98.40% and a recall of over 97.69% in the anchorperson shot detection experiment.

Because of the limit of finite missile time response and the field of view (FOV) of imaging infrared (IR) seeker during the endgame of an intercept, the image of the fighter grows larger gradually and finally will overflow the FOV as the missle approaches the fighter. It then results in losing control of the seeker and affecting the precision of burst control of imaging IR fuze based upon the guidance integrated fuzing (GIF) technology. The aim of the research presented in this paper is to decrease the blind range of imaging IR seeker and improve the precision of aim-point parameters through pose recognition. On the basis of the moving characters of missle and fighter during high-speed encounter and the high correlation of frame to frame in image sequence obtained b imaging IR seeker, a novel method of fighter axis pose recognition and aim-point estimation is proposed. Within this methodology, the spatial pose of fighter axis is recognized before the image overflow the FOV, and then the tracking mode of seeker is transformed from general tracking mode to partial image tracking mode in right time. During partial image tracking, the seeker is controlled to keep the partial image track point in the FOV, then the aim-point parameters can be calculated accurately by utilizing the fighter axis pose, parameters of track point and the relative distance of track point and aim-point.

According to the definition of a curvelet, it is more suitable for image processing than the wavelet and able to represent both smooth and edge parts of image with sparsity. In addition, the representation contains more directional information. The primary experimental results available at present show its potential in image processing. In this paper, we first introduce the concept of curved transform. Then, we propose an image retrieval scheme based on curvelet transform. The experimental results and relevant conclusion are given finally.

Regions of Interest coding (ROI) is the important and convenient image coding technique and is supported by JPEG 2000. It enables a non-uniform distribution of the image quality between selected regions of interest and the background. This paper provides a novel ROI bitplants shift method based on compensation scheme called SHIFT-CS coding algorithm. The method firstly transmits the coding blocks of low-frequency important coefficients that are decoded. The residual coefficients in ROI mask can be transmitted on different bit-planes. When multiple regions of interest are encoded, every ROI has several bit-plane to be transmitted that can be resolved by the ROI's importance. In addition, different objectors can select different regions of interest by the low-quality reconstructive image according to different needs that realize to select ROI based on interactive net browser. The experiments show that SPIFT-CS method combines advantages of both ROI methods. Because the algorithm remain compliant with the MAXSHIFT decoding algorithm described in JPEG 2000 part 1, it is also simple and can be handled by any JPEG 2000 decoder.

In this paper an object recognition method based on genetic algorithm was presented. One of the most difficult problems in object recognition is to correctly decide whether two dissimilar images are originated from different items, or belong to the same object but viewed from different camera positions. Invariant-object recognition is to identiy an object independently of its position (translated or rotated) and size (larger or smaller). It was found that images of objects observed from two different viewpoints can be approximately related by affine transformation if the camera is placed sufficiently far away. This paper proposes an affine and projective invariant ship object recognition approach and employs genetic algorithm (GA) to find the appropriate affine transformation parameters to implement the matching process between the reference object and the scene object. Experiment results show good performances of the proposed method.

This work investigates the image registration from feature point sets. Image registration is a fundamental object recognition method in computer vision and aims to find best matches between two or more point sets when tehre are geometric distortions, point measurement errors and contamination present. This paper concentrates on image registration from feature point sets whose transformation is affine or projective and gives closed form solutions of the transformation parameters, respectively. Furthermore, the Random Sample Consensus algorithm is applied to the matching problem in order to yield the transformation, as well as a set of points in correspondence from images directly. The experimental results show that the methods are accurate, stable and are affected slightly by noise.

In allusion to the features of dynamic multi-target tracking and measuring system (DMTTMS), compars the DMTTMS with the single target tracking and measuring system (STTMS) and analyses the difficulties about homonymy image point ascertainment in DMTTMS. Three methods are presented based on the geometric peculiarity of rays in imaging principia of geometric optics to solve the problem of homonymy image point ascertainment. A design scheme of DMTTMS is put forward using multiple optical capture instruments. Furthermore, an algorithm is emphasized that could treat some targets that are hided by other targets. The simulating result shows that proposed scheme and algorithm has feasibility and validity for DMTTMS.

The linear trajectory is one of major performance for industrial robot. A vision-based robots' linear trajectory measure system is introduced in this paper using a structure light and a special measure track. The three inflexions of the optical strip imaging at the V shape track are used to compute the pose between the sensor frame and the track frame, then the linear trajectory of robot can be computed. The emphasis of this paper is the image processing. At this paper, the process of the image processing method for this system will be described at first, then the key methods include image segmentation and line fitting will be discussed, at last the experiment results will be given.

For the drawbacks of the available algorithms for shot boundary detection of gradual transitions, this paper presents a hierarachical detection method, cutting before detection. This method combines wavelet analysis, fuzzy clustering and statistical techniques, which can distinguish the video shots with gradual transitions from those with abrupt transition. While it can locate the gradual shot boundary accurately. Since the proposed method does not need any empirical threshold, it avoids the interaction with human. The experimental results with real video clips demonstrates its effectiveness and robustness.

In this paper, we propose content-based video retrieval, which is a kind of retrieval by its semantical contents. Because video data is composed of multimodal information streams such as video, auditory and textual streams, we describe a strategy of using multimodal analysis for automatic parsing sports video. The paper first defines the basic structure of sports video database system, and then introduces a new approach that integrates visual stream analysis, speech recognition, speech signal processing and text extraction to realize video retrieval. The experimental results for TV sports video of football games indicate that the multimodal analysis is effective for video retrieval by quickly browsing tree-like video clips or inputting keywords within predefined domain.

Scene change detection is an essential step to automatic and content-based video indexing, retrieval and browsing. In this paper, a robust scene change detection and classification approach is presented, which analyzes audio, visual and textual sources and accounts for their inter-relations and coincidence to semantically identify and classify video scenes. Audio analysis focuses on the segmentation of audio stream into four types of semantic data such as silence, speech, music and environmental sound. Further processing on speech segments aims at locating speaker changes. Video analysis partitions visual stream into shots. Text analysis can provide a supplemental source of clues for scene classification and indexing information. We integrate the video and audio analysis results to identify video scenes and use the text information detected by the video OCR technology or derived from transcripts available to refine scene classification. Results from single source segmentation are in some cases suboptimal. By combining visual, aural features adn the accessorial text information, the scence extraction accuracy is enhanced, and more semantic segmentations are developed. Experimental results are proven to rather promising.

Practical content-based image retrieval systems require efficient indexing schemes for fast k-nearest neighbor (k-NN) searches. Researchers have proposed many tree-based methods using space and data partitioning for similarity searches. However, traditional indexing methods perform poorly and will degrade to simple sequential scans at high dimensionality - that is so-called "curse of dimensionality". Recently, several filtering approaches based on vector approximation (VA) were proposed and showed promising performance. However, VA-based approaches need compute the bound of the distance between each feature vector and the query. It will consume the same computational overhead as the brute-force sequential scan. In this paper, a novel hierarchical indexing scheme is proposed. This approach integrates VA-based index structure with approximate NN (ANN) searches and performs probabilistic ANN searches on approximate vectors. Experiments show the proposed approach achieves a remarkable reduction of computational overhead and disk accesses for k-NN searches. This presented approach supports quadratic-form distance metric and can integrate with relevance feedback techniques for practical large-scale image retrieval systems.

Remote sensing data, especially the hyperspectral remote sensing data, characterize their great quantities. So how to deal wtih these data is a focus. Database has solved the problem of storing, searching, updating and maintaining of the data, but it is not satisfactory in disposing them. In recent years, the technology of data warehouse has great development. It can re-integrate, synthesize and separate the data of database, and use the searching pattern of multiple dimensions to realize data mining (DM). This technolgoy has been widely used in commerce to analyze the inner relationship of the numerous data and makes some remarkable achievements in decision supporting. Data warehouse and Data mining technology have been used in GIS. This article would give a set of complete steps and some general methods in using the DM to analyze the remote sensing data, especially in hyperspectral data. And it tries to do some preliminary exploration in using it to deeply analyze the potential relations among the acquired spectra, images and biology parameters of the experiments and get some anticipated possible results.

A novel contour-based 3D terrain matching method is presented in this paper. In the method, Iso-Elevation Contour Map (IECM), a compact feature-based representation, is proposed to represent the reference DEM and recovered DEM(REM) from real-time data to convert 3D terrain matching to contour-based matching. In the contour-based matching, a normalized wavelet descriptor, which is invariant to 2D rigit transformation, is employed to describe contours. A very fast contour-matching algorithm based on normalized wavelet descriptor is presented. The proposed matching method is robust and effective computation, and can achieve high location accuracy.

A new approach for three-dimensional reconstruction of industrial parts with non-metric image sequences and hybrid point-line photogrammetry is proposed. Non-metric image sequence and CAD-designed data are used as source of information. The strategy of our approach is to reconstruct the parts automatically with points and line segments extracted from imagery. Hybrid point-line photogrammetry is used to reconstruct sheetmetal parts accurately, and the reconstructed model can be used for visualization and inspection. The reconstruction system can run automatically and fastly, and the output of hybrid point-line photogrammetry is the final 3D geometric model of the part. Results of real images of several parts are very satisfying, which shows a promising potential in automatic 3D reconstruction of widely existed industrial parts mainly composed of points and lines.

In this paper, a new approach for automatic creating boundary-representation of planar objects from single line drawings depicting the objects was proposed. Compared with other reconstruction methods that use information a bout points, for computer visiona nd CAD system, it was more convenient to use information about lines to interpret line drawings of polyhedral objects. The coordinates relations between a three-dimensional line and its perspective projection in two-dimension plane are given. Another three coordinates relations are also provided, namely, the coordinates relations when two lines are intersecting, parallel, cross, perpendicular, relations when a line is parallel with or belongs to a plane, and relations when two planes parallel each other. On the basis of these coordinates relations, some new constraints about lines and planes are proposed and a linear system is established. The 3D information of planar objects can be obtained by solving the linear system. In the interpretation of line drawing of a planar object, it is proved that there are at least four degrees of freedom. Some examples are given.

Interest in 3D city Modeling and quickly updating have significantly raised in the past years. In this paper, an economically viable solution for city modeling based on strict digital photogrammetric theory is proposed. Two oblique image sequences for buildings' walls and one vertical image sequence for building's roof, acquired by digital video camera on helicopter, coarse 2D vector data of buildings and LIDAR data are used as sources of information. Camera parameters of each image are acquired by automatic aerial triangulation technique for high overlapping image sequence. By matching every space edge of the building to its image line, the building is accurately reconstructed. The visualization of 3D city model is realized based on the reconstructed buildings with fine textures superimposed on DSM and orthoimage.

Two key important techniques of DTM (Digital Terrain Model) modelling from remote sensing image pair with high resolution (RSIPHR) are introduced in this paper. First is determining conjugate point pair automatically by image matching. Because the approximate epipolar image pair of some RSIPHR remains quite large y-parallelx after relative registration, 2 D relaxation matching sh ould be used in the DTM Modeling. The procedure includes feature point extraction, approximate value estimation, matching and refining, check and filter. Second is the calculation of the spatial coordinates for the conjugate point pair from matching. For IKONS and QUICKBIRD images, it can be based on the RPC/RPB (Rational Polynomial Coefficients) parameters. Because the coordinate accuracy, computed by RPC/RPB parameters, is quite lower in many cases, a block adjustment with an affine transform of image coordinates should be completed based on some control points. Then, the affine transform and RPC/RPB parameters should be used in computation. If there are control points more than 4, the new, simple and strict geometric model based on affine transformation could be applied for any RSIPHR. Only 8 affine coefficients and one slantwise angle need to be determined by control points. Then, they can be applied in the computation.

Multiple digital watermarking is attracting more and more researchers because it is more valuable in the practical applications than single watermarking. In this paper, a multiple watermarking algorithm based on 1-D and 2-D chaotic sequences is proposed. The chaotic sequences have the advantages of massive, high security, and weakest correlation. The massive and independent digital watermark signals are generated through 1-D chaotic maps, which are determined by different initial conditions and parameters. The chaotic digital watermark signals effectively resolve the construction of massive watermarks with good performance. The embedding of multiple watermakrs is more complex than the single watermarking scheme. In this paper, each watermark is added to the middle frequency coefficients of wavelet domain randomly by exploiting 2-D chaotic system, so the embedding and extracting of each watermark would not disturb each other. Considering the parameters of 2-D chaotic systsem as the key to embedding procedure can prevent the watermarks to be removed maliciously, therefore the performance of security is better. The capacity of the multiple watermarking is also analyzed in this paper. The experimental results demonstrate that this proposed watermarking algorithm is robust to many common attacks and it is a reliable copyright protection for multiple legal owners.

Many types of feature extracting of RS image are analyzed, and the work procedure of pattern recognizing in RS images of seismic disaster is proposed. The aerial RS image of Tangshan Great Earthquake is processed, and the digital features of various typical seismic disaster on the RS image is calculated.

Hyperspectral sounder data is a particular class of data that requires high accuracy for useful retrieval of atmospheric temperature and moisture profiles, surface characteristics, cloud properties, and trace gas information. Therefore compression of these data sets is better to be lossless or near lossless. The next-generation NOAA/NESDIS GOES-R hyperspectral sounder, now referred to as the HES (Hyperspectral Environmental Suite), will have hyperspectral resolution (over one thousand channels with spectral widths on the order of 0.5 wavenumber) and high spatial resolution (less than 10 km). Given the large volume of three-dimensional hyperspectral sounder data that will be generated by the HES instrument, the use of robust data compression techniques will be beneficial to data transfer and archive. In this paper, we study lossless data compression for the HES using 3D integer wavelet transforms via the lifting schemes. The wavelet coefficients are then processed with the 3D embedded zerotree wavelet (EZW) algorithm followed by context-based arithmetic coding. We extend the 3D EZW scheme to take on any size of 3D satellite data, each of whose dimensions need not be divisible by 2^N, where N is the levels of the wavelet decomposition being performed. The compression ratios of various kinds of wavelet transforms are presented along with a comparison with the JPEG2000 codec.