This PDF file contains the front matter associated with SPIE Proceedings Volume 7546, including the Title Page, Copyright information, Table of Contents, Introduction, and the Conference Committee listing.

This paper presents a new robust watermarking scheme based on a block probability in wavelet domain. A binary watermark image is permutated by a chaotic function and a secret key. apply Discrete Wavelet Transform (DWT) to decompose the cover host image into four non-overlapping multi resolution sub-bands and then each bit of the binary encoded watermark is embedded by modifying the intensities of a non-overlapping block of 4×4 of the HL sub-band using a probability method. The extraction of the watermark is by comparing the intensities of a block of 4×4 of the watermarked and the original images and calculating the probability of detecting '0' or '1'. Experimental results show that the proposed scheme is robust and secure against a wide range of image processing operations.

The idea of this project development is to improve the concept of human face recognition that has been studied in order to apply it for a more precise and effective recognition of human faces, and offered an alternative to agencies with respect to their access-departure control system. To accomplish this, a technique of calculation of distances between face features, including efficient face recognition though a neural network, is used. The system uses a technique of image processing consisting of 3 major processes: 1) preprocessing or preparation of images, 2) feature extraction from images of eyes, ears, nose and mouth, used for a calculation of Euclidean distances between each organ; and 3) face recognition using a neural network method. Based on the experimental results from reading image of a total of 200 images from 100 human faces, the system can correctly recognize 96 % with average access time of 3.304 sec per image.

In the research we used "Method of Constant Stimuli". Stimuli were presented numerous times in random order and the subject reports whether he/she could detect them. In Stage 1, we found the speed smudges disappear is Ratio 10:16 > Ratio 10:14 > Ratio 10:12 > Ratio 10:12, no matter when we right declined the squares or when we rotated the squares. In Stage 2, we found when the style of squares is right declined, the speed smudges disappear was Ratio 10:12 > Ratio 10:11; when the style of squares was down declined, the speed smudges disappear was Ratio 10:11 > Ratio 10:12. In Stage 3 we found that no matter rotated or right declined or down declined or rotated the squares, the speed smudges disappeared was Ratio 10:11 < Ratio 10:10 and Ratio 10:12. Then we can find the effect trend ratio on stopping Hermann Grid may be a (symbol--see manuscript) mark. It means the speed smudges disappear is Ratio 10:16 > Ratio 10:14 > Ratio 10:12 > Ratio 10:10 > Ratio 10:11.

Degraded character recognition is a challenging problem in the field of Optical Character Recognition (OCR). The performance of an optical character recognition depends upon printed quality of the input documents. Many OCRs have been designed which correctly identifies the fine printed documents. But, very few reported work has been found on the recognition of the degraded documents. The efficiency of the OCRs system decreases if the input image is degraded. In this paper, a novel approach based on gradient pattern for recognizing degraded printed character is proposed. The approach makes use of gradient pattern of an individual character for recognition. Experiments were conducted on character image that is either digitally written or a degraded character extracted from historical documents and the results are found to be satisfactory.

A robot navigating in an unknown environment depends on its sensors to obtain distance information of the obstacles with which it encounters. By using the collected distance information, a point robot is able to construct Generalized Local Voronoi Diagram (GLVD) of the area that is lying in vicinity of itself. In this paper, a method is proposed to build GLVDs via processing the digital images captured by a camera attached to a point robot which remains on a constant height platform. The robot under consideration collects several discrete images around itself. These images are processed by an algorithm which extracts the distance information of the three dimensional obstacles around the robot and constructs discrete images which are parts of the top view of the area. The resulting discrete images are then assembled to build a complete image of the top view of the region around the robot, at the center of which the robot is located. This local panoramic image is then used to construct the GLVD of the area under consideration. A case study is presented to demonstrate and verify the capabilities of the approach introduced. To the best of authors' knowledge, this is the first study which employs a camera attached to the robot itself and digital image processing to create Generalized Local Voronoi Diagrams.

Polymorphic robotic systems, which are composed of many modular robots that act in coordination to achieve a goal defined on the system level, have been drawing attention of industrial and research communities since they bring additional flexibility in many applications. This paper introduces a new polymorphic robotic system, in which the detection and control of the modules are attained by a stationary observing camera. The modules do not have any sensory equipment for positioning or detecting each other. They are self-powered, geared with means of wireless communication and locking mechanisms, and are marked to enable the image processing algorithm detect the position and orientation of each of them in a two dimensional space. Since the system does not depend on the modules for positioning and commanding others, in a circumstance where one or more of the modules malfunction, the system will be able to continue operating with the rest of the modules. Moreover, to enhance the compatibility and robustness of the system under different illumination conditions, stationary reference markers are employed together with global positioning markers, and an adaptive filtering parameter decision methodology is enclosed. To the best of authors' knowledge, this is the first study to introduce a remote camera observer to control modules of a polymorphic robotic system.

Image segmentation is one of the significant digital image processing techniques commonly used in the medical field. One of the specific applications is tumor detection in abnormal Magnetic Resonance (MR) brain images. Fuzzy approaches are widely preferred for tumor segmentation which generally yields superior results in terms of accuracy. But most of the fuzzy algorithms suffer from the drawback of slow convergence rate which makes the system practically non-feasible. In this work, the application of modified Fuzzy C-means (FCM) algorithm to tackle the convergence problem is explored in the context of brain image segmentation. This modified FCM algorithm employs the concept of quantization to improve the convergence rate besides yielding excellent segmentation efficiency. This algorithm is experimented on real time abnormal MR brain images collected from the radiologists. A comprehensive feature vector is extracted from these images and used for the segmentation technique. An extensive feature selection process is performed which reduces the convergence time period and improve the segmentation efficiency. After segmentation, the tumor portion is extracted from the segmented image. Comparative analysis in terms of segmentation efficiency and convergence rate is performed between the conventional FCM and the modified FCM. Experimental results show superior results for the modified FCM algorithm in terms of the performance measures. Thus, this work highlights the application of the modified algorithm for brain tumor detection in abnormal MR brain images.

Diabetic retinopathy (DR) is a chronic eye disease for which early detection is highly essential to avoid any fatal results. Image processing of retinal images emerge as a feasible tool for this early diagnosis. Digital image processing techniques involve image classification which is a significant technique to detect the abnormality in the eye. Various automated classification systems have been developed in the recent years but most of them lack high classification accuracy. Artificial neural networks are the widely preferred artificial intelligence technique since it yields superior results in terms of classification accuracy. In this work, Radial Basis function (RBF) neural network based bi-level classification system is proposed to differentiate abnormal DR Images and normal retinal images. The results are analyzed in terms of classification accuracy, sensitivity and specificity. A comparative analysis is performed with the results of the probabilistic classifier namely Bayesian classifier to show the superior nature of neural classifier. Experimental results show promising results for the neural classifier in terms of the performance measures.

This research applied the Euclidean distance technique to generate a system of Thai handwritten character recognition. The system consists of four main components which include: 1) Image Acquisition, 2) Image Pre-processing, 3) Recognition, and 4) Display Result. All training and testing handwritten characters in this research used all Thai native people to write them for avoiding invalid shape of Thai character. The character images fed to the training part totaling 3,513 characters. Out of 878 Thai handwritten characters tested, it was found that the system could recognize (accept) 716 characters or 81.55%, while rejecting 61 characters or 6.95% and misrecognizing 101 characters or 11.50%. We tested the system with 50 Japanese handwritten characters and 25 invalid Thai handwritten character shape, it was found that the system could reject 47 characters or 62.67% while misrecognizing 28 characters or 37.33%.

This paper proposed the discrete transform and neural network algorithms to obtain the monthly peak load demand in mid term load forecasting. The mother wavelet daubechies2 (db2) is employed to decomposed, high pass filter and low pass filter signals from the original signal before using feed forward back propagation neural network to determine the forecasting results. The historical data records in 1997-2007 of Electricity Generating Authority of Thailand (EGAT) is used as reference. In this study, historical information of peak load demand(MW), mean temperature(Tmean), consumer price index (CPI), and industrial index (economic:IDI) are used as feature inputs of the network. The experimental results show that the Mean Absolute Percentage Error (MAPE) is approximately 4.32%. This forecasting results can be used for fuel planning and unit commitment of the power system in the future.

It is necessary for blinding digital image watermark algorithm to extract watermark information without any extra information except the watermarked image itself. But most of the current blinding watermark algorithms have the same disadvantage: besides the watermarked image, they also need the size and other information about the original image when extracting the watermark. This paper presents an innovative blinding color image watermark algorithm based on Lab color space, which does not have the disadvantages mentioned above. This algorithm first marks the watermark region size and position through embedding some regular blocks called anchor points in image spatial domain, and then embeds the watermark into the image. In doing so, the watermark information can be easily extracted after doing cropping and scale change to the image. Experimental results show that the algorithm is particularly robust against the color adjusting and geometry transformation. This algorithm has already been used in a copyright protecting project and works very well.

This paper presents a framework for robustly and accurately computing the visual hull of a real object from images sequences. Unlike most existing volumetric based approaches, level set deformable model is utilized in our system to drive the surface from a sphere smoothly recovery the shape of the real object. The algorithm represents the object's surface implicitly as the zero level set in uniform grid and the visual hull computation problem is translated into a forces computation problem. The deforming surface evolves under the internal and external forces according to the silhouettes and smoothness constrains. Snake deformable model is applied as a refinement step to improve the quality of mesh and reduce the total computing time. This classical and geometric mixed deformation model can easily and naturally changes the topology of the surface and need not add any extra measurement to avoid mesh confusion. The experiment results turns out that the final mesh have higher precise and smoothness than the traditional volumetric based approaches.

Finding a best clustering algorithm to tackle the problem of finding the optimal partition of a data set is always an NP-hard problem. In general, solutions to the NP-hard problems involve searches through vast spaces of possible solutions and evolutionary algorithms have been a success. In this paper, we explore one such approach which is hardly known outside the search heuristic field - the Particle Swarm Optimisation+k-means (PSOk) for this purpose. The proposed hybrid algorithm consists of two modules, the PSO module and the k-means module. For the initial stage, the PSO module is executed for a short period to search for the clusters centroid locations. Succeeding to the PSO module is the refining stage where the detected locations are transferred to the k-means module for refinement and generation of the final optimal clustering solution. Experimental results on two challenging datasets and a comparison with other hybrid PSO methods has demonstrated and validated the effectiveness of the proposed solution in terms of precision and computational complexity.

A personal identification system using finger-vein patterns and neural network techniques is proposed in the present study. In the proposed system, the finger-vein patterns are captured by a device that can transmit near infrared through the finger and record the patterns for signal analysis and classification. The biometric system for verification consists of a combination of feature extraction using principal component analysis and pattern classification using both back-propagation network and adaptive neuro-fuzzy inference systems. Finger-vein features are first extracted by principal component analysis method to reduce the computational burden and removes noise residing in the discarded dimensions. The features are then used in pattern classification and identification. To verify the effect of the proposed adaptive neuro-fuzzy inference system in the pattern classification, the back-propagation network is compared with the proposed system. The experimental results indicated the proposed system using adaptive neuro-fuzzy inference system demonstrated a better performance than the back-propagation network for personal identification using the finger-vein patterns.

The development of digital video technology, due to its nature, introduced new approach to the objective video quality estimation. Basically there are two types of metrics for measuring the quality of digital video: purely mathematically defined video quality metrics (DELTA, MSAD, MSE, SNR and PSNR) where the error is mathematically calculated as a difference between the original and processed pixel, and video quality metrics that have similar characteristics as the Human Visual System (SSIM, NQI, VQM), where the perceptual quality is also considered in the overall quality estimation. The metrics from the first group are more technical ones and because the visual quality of perception is more complex than pixel error calculation, many examples show that their video quality estimation is deficiently accurate. The second group of metrics work in a different manner compared to previous, calculating the scene structure in the overall video quality estimation. This paper is concerned with experimental comparison of the performance of Structural Similarity (SSIM) and Video Quality Metric (VQM) metrics for objective video quality estimation. For the purpose of this experiment, more than 300 short video sequences were prepared. The measurements of these video sequences are used to draw the metrics dependence to common changes in processed video sequences. These changes include changes in: brightness, contrast, hue, saturation and noise. This paper pinpoints the key characteristics of each metric, gives the conclusion of the better performing one and gives directions for improvement of objective video quality estimation.

Using 3D visualization technology, environment and objects in the nature can be represented in a dynamic and intuitional way. It improves the effect of perceiving terrain and environment, which is a current research trend of geographic information system. Through the analysis of 3DGIS and related visualization technology, the paper concluded the research status of 3DGIS related technologies. Based on the research of 3D terrain, terrain features and real time display technologies, the 3DGIS visualization related technologies were concluded. The future development orientation of 3DGIS technology was also analyzed in the paper. The research concludes related works and orients the future work of 3DGIS technologies.

Extraction of elliptic shapes in real images is very challenging because the geometric shapes corresponding to the various objects often appear incomplete and deformed due to the presence of noise, cluttered background and occlusion by other objects. This paper proposes a new method of ellipse detection, which is able to deal with the challenges mentioned above, while being computationally efficient and more accurate than existing methods. The novelty of the current work is a grouping scheme based on a 'trust score' that indicates the trust that can be put upon an edge in a group. In the first stage, partial Hough transform is performed in order to generate the possible centers (or center bins in 2-dimensional pixel space). Then, a special histogram is generated using the 'trust score' that rates the relationship of the edge and the center bin. This histogram is used to group the edges and rank them within each group. In the second stage, least square technique is applied in order to judge and improve the grouping and finally find the parameters of the ellipses. Such hybrid method has various advantages like consideration of large number of possible groups, computational efficiency, parallelizability, real time application, etc. The method performs well for complicated real images and is suitable for real-time applications of machine vision.

In this paper, a novel algorithm using vector quantization (VQ) method for facial image recognition in DCT domain is presented. Firstly, feature vectors of facial image are generated by using DCT (Discrete Cosine transform) coefficients in low frequency domains. Then codevector referred count histogram, which is utilized as a very effective personal feature value, is obtained by Vector Quantization (VQ) processing. Publicly available AT&T database of 40 subjects with 10 images per subject containing variations in lighting, posing, and expressions, is used to evaluate the performance of the proposed algorithm. Experimental results show face recognition using proposed feature vector is very efficient. The highest average recognition rate of 94.8% is obtained.

In this paper, a fast search algorithm for MPEG-4 video clips from video database is proposed. An adjacent pixel intensity difference quantization (APIDQ) histogram is utilized as the feature vector of VOP (video object plane), which had been reliably applied to human face recognition previously. Instead of fully decompressed video sequence, partially decoded data, namely DC sequence of the video object are extracted from the video sequence. Combined with active search, a temporal pruning algorithm, fast and robust video search can be realized. The proposed search algorithm has been evaluated by total 15 hours of video contained of TV programs such as drama, talk, news, etc. to search for given 200 MPEG-4 video clips which each length is 15 seconds. Experimental results show the proposed algorithm can detect the similar video clip in merely 80ms, and Equal Error Rate (ERR) of 2 % in drama and news categories are achieved, which are more accurately and robust than conventional fast video search algorithm.

Generating virtual face images with different poses has potential applications in many areas, such as face recognition, human-machine interaction, portrait combination, and computer graphics. However, in some situation, the available face images are quite limited, which makes the problem difficult. This paper proposes a pose-specific shape eigenspace based face wrapping method to generate virtual face images with different poses from a specific pose. A predefined training set is necessary. According to their poses, training faces with annotated landmarks are manually divided into several groups, each of which is utilized to learn a pose-specific shape eigenspace by K-L transform. For a new image under a certain pose, its shape information described by the annotated landmarks is firstly projected to the expected pose-specific shape eigenspace to represent the shape information of this image under the expected pose. Then, all corresponding points between the represented shape and original shape the are matched and the texture information of all points in the represented shape are covered by the gray or color information of the corresponding points in the original image to generate a virtual face image under expected pose. To quantify the similarity between the generated virtual images and real images, cosine similarity is adopted. Experiments on IMM, PIE and YaleB face subsets show that the similarity of the virtual image and real images is over 0.9, no matter there is high or low similarity between test set and training set, which illustrates the effectiveness of the proposed method.

Motivated by the mechanisms of mammalian primary visual cortex (V1), we propose a hierarchical model of feature extraction for object recognition. The proposed model consists of two layers, each of which emulates the functions of V1 simple cells and complex cells respectively. Filters learned from training images are applied at every position of the input image to get an edge feature representation. Then a maximum pooling operation is taken to increase shiftinvariance of the feature. Experiments on face recognition and crop-wise object detection show that our model is competitive with the state-of-the-art biologically-inspired method.

Support vector machine (SVM) has been proved to be a powerful tool for face recognition. The generalization capacity of SVM depends on the model with optimal hyperparameters. The computational cost of SVM model selection results in application difficulty in face recognition. In order to overcome the shortcoming, we utilize the advantage of uniform design--space filling designs and uniformly scattering theory to seek for optimal SVM hyperparameters. Then we propose a face recognition scheme based on SVM with optimal model which obtained by replacing the grid and gradient-based method with uniform design. The experimental results on Yale and PIE face databases show that the proposed method significantly improves the efficiency of SVM model selection.

Image similarity measurement is one of the key research areas in pattern recognition and image retrieval. The paper presents a new, clustering based approach, called Image Co-Clustering. The term co-clustering is related to cosegmentation, where two or more images are segmented simultaneously. The proposed approach simultaneously clusters feature vector sets from two images. Generated clusters represent similar parts in both images. To have visually coherent clusters, spatial constrains should be imposed onto the clustering algorithm. The key difficulty is to impose spatial constrains in such a way, that the same objects having different locations, rotations and scales on two images are captured in the same cluster. The proposed spatial constraining is called partial spatial constraining.

The identification of human basic actions plays an important role for recognizing human activities in complex scene. In this paper we propose an approach for automatic human action recognition. The parametric model of human is extracted from image sequences using motion/texture based human detection and tracking. Action features from its model are carefully defined into the action interaction representation and used for the recognizing process. Performance of proposed method is tested experimentally using datasets under indoor environments.

Due to the genetic proximities, siblings are often observed to bear close facial resemblances to each other as well as their parents. In this paper, we attempt to develop such human capability in computers. In order to achieve this goal, Haar, Gabor, SIFT and SURF features of family and nonfamily datasets are extracted and used for AdaBoost to train the classifier. The primary difference between our study and other relevant applications like face recognition, album auto tagging and annotation is that the query person we intend to classify may not even exist in the training data. We have conducted testing for various scenarios where different members of the family are absent from training but present in testing, and have obtained interesting results with practical implications for the development of automated family member classification. As family data sets used in this paper has good quality colour samples, we use FERET dataset as non-family samples to have fair comparison. Results obtained show that we can achieve up to 87% accuracy depending on the absent family member.

This paper present the classification of fresh aromatic coconuts into 3 types: single layer, double layer and one and a half layer by inspecting colors at the bottom of coconuts. We take the photos the bottom of coconuts in RGB mode, change the colors into the HSV mode, and then place 4 circles into the image. The 20 photos of each type are used to generate the relation of the rings for each type by using polynomial regression. Finally, we use the polynomial equations to test new 100 fresh aromatic coconuts, the result is 11.76% errors for single layer, 18.6% for one and a half layer and 18.18% error for double layers.

Automatic face detection in colored images is closely related to face recognition systems, as a preliminary critical required step, where it is necessary to search for the precise face location. We propose a reliable approach for skin color segmentation to detect human face in colored images under unconstrained scene conditions that overcoming the sensitivity to the variation in face size, pose, location, lighting conditions, and complex background. Our approach is based on building multi skin color clustering models using HSV color space, multi-level segmentation, and rule-based classifier. We proposed to use four skin color clustering models instead of single skin clustering model, namely: standard-skin model, shadow-skin model, light-skin model, high-red-skin model. We made an independent skin color clustering models by converting 3-D color space to 2-D without losing color information in order to find the classification boundaries for each skin color pattern class in 2-D. Once we find the classification boundaries, we process the input image with the first-level skin-color segmentation to produce four layers; each layer reflecting its skin-color clustering model. Then an iterative rule-based region grow is performed to create one solid region of interest which is presumed to be a face candidate region that will be passed to the second-level segmentation. In this approach we combine pixel-based segmentation and region-based segmentation using the four skin layers. We also propose skin-color correction (skin lighting) at shadow-skin layer to improve detection rate. In the second-level segmentation we use gray scale to segment the face candidate region into the most significant features using thresholding. Next step is to compute the X-Y-reliefs to locate the accurate position of facial features in each face candidate region and match it with our geometrical knowledge in order to classify the face candidate region to a face or non-face region. We present experimental results of our implementation and demonstrate the feasibility of our approach to be general purpose skin color segmentation for face detection problem.

Granular data classification and clustering is an upcoming and important issue in the field of pattern recognition. This paper proposes a Supervised Colour Image Segmentation technique based on Granular Reflex Fuzzy Min-Max Neural Network (GrRFMN). GrRFMN architecture consists of a reflex mechanism inspired from human brain to handle class overlaps. It has been observed that most of the image segmentation techniques are pixel based. It means that segmentation is done on pixel-by-pixel basis. In this paper, a novel granule based approached for colour image segmentation is proposed. In the proposed technique granules of an image are processed. This results into a fast segmentation process. The image segmentation discussed here is a supervised. In training phase, GrRFMN learns different classes in the image using class granules. A trained GrRFMN is then used to segment the image. As GrRMN is trainable on-line in a single pass through data, the proposed method is easily extended for video sequence segmentation. Results on various standard images are presented.

In this paper, novel method based on Dual Tree Complex Wavelet Packet Transform (DT-CWPT) to analyze texture of Computer Tomography (CT) images and extract the corresponding feature vectors for content based medical image retrieval is proposed. This is mainly because of DT-CWPT characterizes textural property of CT images in better way. The feature vectors of CT images are extracted by measuring energy and standard deviation of DT-CWPT subband. These features are used to facilitate content based medical image retrieval (CBMIR).The proposed method outperforms than existing available methods.

This paper presents an application of discrete cosine transform compression technique on color medical images. The binary matrix technique is used to encode the coefficients. Initially the image is divided into R,G,B planes, discrete cosine transform(DCT), threshold function, quantization and binary matrix technique is applied to the planes. In the binary matrix technique the highest probable gray level is coded as zero in binary matrix and other gray levels with one. The most probable gray level is stored as first element in the array followed by other gray levels in the order as they appear in the image. The image is represented as binary matrix and gray level array. During decompression phase the binary matrix acts as look up table providing values for the reconstructed matrix. Since most of the values in individual planes are same after quantization this procedure yields good compression ratio. The tests of this lossy compression/decompression technique are performed on medical images, the obtained results Figure [3-11] shows that the DCT technique permits to considerably improve the compression rate while maintaining a good image quality when threshold varies in the interval: 0 ≤ TH ≤ 20 Q=8 for block sizes:[4×4]and [8×8].The computational complexity is greatly reduced thus producing faster compressions and decompressions. Due to the combination of simplicity, compression potential and image quality the proposed algorithm is best suitable for medical images.

License plate detection and recognition are vital yet challenging tasks for law enforcement agencies. This paper presents a license plate detection prototype system for a Macao law enforcement department using Modest Adaboost combined with template matching technique. Firstly, a machine learning algorithm, based on Modest AdaBoost which mostly aims for better generalization capability and resistance to overfitting, was applied to find out candidate license plates over the input images. In the second stage, template matching technique was employed to verify the license plate appearances in order to reduce false positives. This paper shows that the AdaBoost algorithm, which was originally used for face detection, has successfully been applied to solve the problems of license plate detection. Experimental results demonstrate high accuracy and efficiency of the proposed method.

The present paper aims to introduce a new method of face recognition based on integrating the results of three different neural networks and discuss the final outcome from a fuzzy point of view (recognition classifier). The first merit of this method is that it is not relying on the positions of eyes and lip on an individual's face. The second is that even if the face is partially covered, the method appears fault tolerant. All the experiments of the study were carried out based on the ORL (Olivetti Research Laboratory) database with 5 training images. For the selected numbers of 20, 30, and 40 subjects, we came to the results of 94%, 92.5%, and 90.25% respectively.

Person-specific image searching and retrieval is an important issue in several areas, including biometrics, robot vision, human-computer interfaces and surveillance. A wildly accepted retrieval methods are always relevant with either large-scale features description or complicated classifiers design. In this paper a system using an image clustering method is presented, which enables fast approximate search based on person face image. First, for face detection, both skin color segmentation strategy and the AdaBoost algorithm have been employed. In clustering, different image streams have been achieved in unsupervised manner where no prior knowledge about the input sequence is required. The proposed system applied to a variety of image datasets with satisfactory performance was demonstrated by the experimental results. The proposed method is also highly efficient, since most computations can be out-sourced to the GPU and competitive with other systems presented recently in the literatures.

Anatomical structure tracing on cephalograms is a significant way to obtain cephalometric analysis. Computerized cephalometric analysis involves both manual and automatic approaches. The manual approach is limited in accuracy and repeatability. In this paper we have attempted to develop and test a novel method for automatic localization of craniofacial structure based on the detected edges on the region of interest. According to the grey scale feature at the different region of the cephalometric images, an algorithm for obtaining tissue contour is put forward. Using edge detection with specific threshold an improved bidirectional contour tracing approach is proposed by an interactive selection of the starting edge pixels, the tracking process searches repetitively for an edge pixel at the neighborhood of previously searched edge pixel to segment images, and then craniofacial structures are obtained. The effectiveness of the algorithm is demonstrated by the preliminary experimental results obtained with the proposed method.

This paper presents an approach to reconstruct solid models from triangular meshes of STL files. First, suitable slicing planes should be selected for extracting parallel intersection contours, which will be used for solid model reconstruction. Usually, a suitable flat region of triangular meshes of the STL model is selected as the bottom surface, and it can be fitted into a plane from the selected flat region. The flat region is separated by a mesh segmentation method, which uses a specified small threshold dihedral angle to divide all triangular facets into separated regions. Next, a series of parallel slicing contours are obtained by cutting the STL model through specified parallel cutting planes. Slicing contours are originally composed of a lot of line segments, which should be simplified and refitted into 2D NURBS curves for data reduction and contour smoothing. The number of points on each slicing contour is reduced by comparing the variation of included angles of each two adjacent line segments. Reduced points of each slicing contour are fitted into a NURBS curve in commercial CAD software. Finally, with a series of parallel 2D NURBS curves, the solid model of the STL facets is established by loft operations supplied in almost all popular CAD software. The established solid model can be used for other post processing such as finite element mesh generation.

An adaptive regularization algorithm that combines elementwise photon absorption and data misfit is proposed to stabilize the non-linear ill-posed inverse problem. The diffuse photon distribution is low near the target compared to the normal region. A Hessian is proposed based on light and tissue interaction, and is estimated using adjoint method by distributing the sources inside the discretized domain. As iteration progresses, the photon absorption near the inhomogeneity becomes high and carries more weightage to the regularization matrix. The domain's interior photon absorption and misfit based adaptive regularization method improves quality of the reconstructed Diffuse Optical Tomographic images.

Medical Image databases are a key component in future diagnosis and preventive medicine. Automatic categorization of medical images plays an important role for structuring of given medical databases as well as for searching and retrieval of medical images. This paper focuses on a general framework for efficient representation and classification of X-ray images, appropriate for medical image archives. The proposed methodology is comprised of a graph theoretic image representation scheme and image matching measures. In this work, x-ray images are represented by undirected graphs and categorization is done based on an inexact graph matching scheme, graph edit distance. Initially, an unsupervised clustering algorithm is applied on input x-ray images in order to extract coherent regions in feature space, and corresponding coherent segments in the image content. The segmented images are then represented as graphs, which are used in the image matching process. Finally, the experimental results have also been presented at the end of the paper.

The objective of the present study is to develop an automatic tool to identify and classify the bacterial cells in digital microscopic cell images. Geometric features are used to identify the different types of bacterial cells, namely, bacilli, cocci and spiral. The current methods rely on the subjective reading of profiles by a human expert based on the various manual staining methods. In this paper, we propose a method for bacterial classification by segmenting digital bacterial cell images and extracting geometric features for cell classification. The experimental results are compared with the manual results obtained by the microbiology expert and demonstrate the efficacy of the proposed method.

This paper presents a System-on-Chip (SoC) visual-based driver drowsiness detection system. The system is able to promptly detect the onset of driver drowsiness by monitoring in real-time the accumulated driver's PERCLOS, i.e. proportion of time driver's eyes are closed in a 1-minute interval through non-intrusive camera(s). FPGA hardware is used as its processing platform along with Viola-Jones object detection algorithm. Viola-Jones algorithm uses Haar-like features along with AdaBoost algorithm to achieve good detection performance.

In this paper, we propose a robust multi-layer texture representation for facial expressions. Our representation is built up using multi-resolution (MR) uniform local binary pattern (ULBP) features on multi-areas (MA) in facial image. Experiments show that this multi-resolution and multi-area (MRMA) strategy could both greatly improve the discriminative ability of texture representation. Based on the proposed MRMA ULBP representation for facial expression, we propose a MRMA ULBP representation + SVM classifier facial expression recognition system. Experiments based on 21 trained one-against-one SVM classifiers show average recognition accuracy of 92.59% on JAFFE database.

Sequential Monte Carlo (SMC) simulations are widely used to solve problems associated with complex probability distribution. Intensive computations are their main drawbacks,whic h restrict to be applied to real time applications,a nd thus efficient parallelism under high performance computing environment is crucial to effective implementations,esp ecially for intelligent computer vision systems. The combination of auxiliary variables importance sampling with Markov Chain Monte Carlo (MCMC) resampling for pipelining data are proposed in this paper so as to minimize executive time,whilst improve the estimation accuracy. Experimental resultion a network of workstations composed of simple off-the-shelf hardware components show that the hybrid parallel scheme provides a bottleneck free to reduce executive time with increasing particles,co mpared to the conventional SMC and MCMC based parallel schemes.

Joint registration and segmentation (JRS) is an effective approach to combine the complementary information of segmentation labels with registration parameters. While most such integrated approaches have been tested on static images, in this work we focus on JRS of dynamic image sequences. For dynamic contrast enhanced images, previous works have focused on multi-stage approaches that interleave registration and segmentation. We propose a Markov random field (MRF) based solution which uses saliency, intensity, edge orientation and segmentation labels for JRS of renal perfusion images. An expectation-maximization (EM) framework is used where the entire image sequence is first registered followed by updating the segmentation labels. Experiments on real patient datasets exhibiting elastic deformations demonstrate the effectiveness of our MRF-based JRS approach.

This paper presents a brain tumor segmentation method which automatically segments tumors from human brain MRI image volume. The presented model is based on the symmetry of human brain and level set method. Firstly, the midsagittal plane of an MRI volume is searched, the slices with potential tumor of the volume are checked out according to their symmetries, and an initial boundary of the tumor in the slice, in which the tumor is in the largest size, is determined meanwhile by watershed and morphological algorithms; Secondly, the level set method is applied to the initial boundary to drive the curve evolving and stopping to the appropriate tumor boundary; Lastly, the tumor boundary is projected one by one to its adjacent slices as initial boundaries through the volume for the whole tumor. The experiment results are compared with hand tracking of the expert and show relatively good accordance between both.

Human beings are commonly identified by biometric schemes which are concerned with identifying individuals by their unique physical characteristics. The use of passwords and personal identification numbers for detecting humans are being used for years now. Disadvantages of these schemes are that someone else may use them or can easily be forgotten. Keeping in view of these problems, biometrics approaches such as face recognition, fingerprint, iris/retina and voice recognition have been developed which provide a far better solution when identifying individuals. A number of methods have been developed for face recognition. This paper illustrates employment of Gabor filters for extracting facial features by constructing a sliding window frame. Classification is done by assigning class label to the unknown image that has maximum features similar to the image stored in the database of that class. The proposed system gives a recognition rate of 96% which is better than many of the similar techniques being used for face recognition.

In this paper, we proposed a basis-background subtraction method using non-negative matrix factorization (NMF). The core idea is to learn the parts of complex background environments by NMF algorithm and exploit the discrimination information in the training set to boost the reconstruction capability of the background efficiently. The method utilize the distance between an observed image and the reconstructed background image for segmenting foreground objects. The principle component analysis (PCA) is used for the enhanced initialization of NMF algorithm. A kind of off-line basis-background maintenance scheme is introduced instead of an incremental learning. A variety of experiments are conducted and illustrate the effectiveness in background subtraction. Quantitative evaluation and comparison with the existing methods show that the proposed method provides good improved results.

Feature extraction using linguistic abstracts described by field experts, and their pragmatic behavior when tested with an inference engine is of interest to computer vision researchers. Advances in image processing have added to the complexity involved with selecting an appropriate feature extraction method for describing a linguistic feature. In this work, we propose the usage of a set of statistical tools for evaluating the efficacy of a feature extraction technique suitable for expressing a linguistic feature. This set of tools are based on expression of class discrimination strength of features, overlap in their expression, and the density of outliers present in them. The feature extraction techniques are ranked based on the scores obtained by them when tested with these tools. An experimental study for validating these claims, based on classification of two different visual texture, expressed using six different texture quantification techniques is also presented.

In this paper, we propose a new approach for skin detection in images taken of different people under various illumination conditions utilizing colors and image segmentation based on edge and region integration. The algorithm incorporates vector-based color edge detection, color quantization, and a new kind of region growing. We achieve satisfactory results that most skin areas are detected correctly and efficiently. Our main contribution lies in the combination of multiple cues and fusion of skin detection and image segmentation.

This paper presents a novel hybrid method for hand segmentation in color imagery. In order to overwhelm such complex details of the color image, the system combines artificial intelligence techniques to achieve automatic hand segmentation. These techniques include self-organizing map, backpropagation artificial neural network, genetic algorithm, convex and deformable template techniques. The proposed system can detect hands without using any initial conditions and can also perform in different light conditions. Moreover, the system can detect hands in different sizes and orientations. This work utilizes convex and deformable template techniques together which allow the system to detect hands at the maximum accuracy of 98 percent.

Here we present an approach of meaningful curve identification with its depth estimation by chaining of the edge points, to locate and track the obstacles with stereo matching for automatic vehicle navigation. We use a self adoptive and nonlinear principle of extended declivity to obtain the edge points (horizontal declivities) in the images. These edge points include lots of noise and hence matching is not effective directly. The large size of the matching problem does not allow us to use effective matching algorithm properly. We use basic assumptions of continuity in the shape of expected obstacles to reduce the problem size and match less number of features effectively. Vertical chaining is used to obtain features which can be used for the tracking or stereo and obtain obstacles in the region of interest. These newly proposed curves are defined with their features and a matching algorithm is used to obtain results.

This paper presents a robust algorithm for digital image watermarking based on Human Visual System (HVS). Watermark is embedded in the Slant Transform domain by altering the transform coefficients. The perceptibility of the watermarked image using proposed algorithm is improved over DCT based algorithm⁹ by embedding the watermark image in selected positions based on the HVS weightage matrix. The proposed method is robust and the watermark image can survive to many image attacks like noise, bit plane removal, cropping, histogram equalization, rotation, and sharpening. Results are compared with DCT based watermarking method and found to be superior in terms of the quality of the watermarked image and resilience to attacks. The metrics used to test the robustness of the proposed algorithm are Peak Signal to Noise Ratio (PSNR) and Normalized Cross Correlation (NCC).

This paper deals with improved ECG signal analysis using Wavelet Transform Techniques and employing subsequent modified feature extraction for Arrhythmia detection based on Neuro-Fuzzy technique. This improvement is based on suitable choice of features in evaluating and predicting life threatening Ventricular Arrhythmia . Analyzing electrocardiographic signals (ECG) includes not only inspection of P, QRS and T waves, but also the causal relations they have and the temporal sequences they build within long observation periods. Wavelet-transform is used for effective feature extraction and Adaptive Neuro-Fuzzy Inference System (ANFIS) is considered for the classifier model. In a first step, QRS complexes are detected. Then, each QRS is delineated by detecting and identifying the peaks of the individual waves, as well as the complex onset and end. Finally, the determination of P and T wave peaks, onsets and ends is performed. We evaluated the algorithm on several manually annotated databases, such as MIT-BIH Arrhythmia and CSE databases, developed for validation purposes. Features based on the ECG waveform shape and heart beat intervals are used as inputs to the classifiers. The performance of the ANFIS model is evaluated in terms of training performance and classification accuracies and the results confirmed that the proposed ANFIS model has potential in classifying the ECG signals. Cross validation is used to measure the classifier performance. A testing classification accuracy of 95.13% is achieved which is a significant improvement.

This paper presents an enhancement method based on human visual characterstics (HVC) for medical images. In medical field images suffer from poor contrast and sometimes information is hidden in dark areas, due to this we are not able to extract information from them. We are presenting a method which takes care of these factors. According to HVC, human eyes are more sensitive towards plenty of details or great changings and less sensitive to smooth regions. So we divide the images in smooth area and detail area by discrete wavelet transform (DWT), and then use different processing methods for these areas according to HVC. Moreover, our experimental results validate that the proposed method performs better than conventional histrogram equalization method.

In this paper, a robust image watermarking method in two-dimensional space/spatial-frequency distributions domain is proposed which is robust against geometric distortion. This watermarking is detected by a linear frequency change. The one-dimensional S transformation is used to detect the watermark. The chirp signals are used as watermarks and this type of signals is resistant to all stationary filtering methods and exhibits geometrical symmetry. In the two-dimensional Radon-Wigner transformation domain, the chirp signals used as watermarks change only its position in space/spatialfrequency distribution, after applying linear geometrical attack, such as scale rotation and cropping. But the two-dimensional Radon-Wigner transformation needs too much difficult computing. So the image is put into a series of 1D signals by choosing scalable local time windows. The watermark embedded in the 1D S transformation domain. The watermark thus generated is invisible and performs well in StirMark test and is robust to geometrical attacks. Compared with other watermarking algorithms, this algorithm is more robust, especially against geometric distortion, while having excellent frequency properties.

This paper presents a new interpolation algorithm based on the adaptive 2-D autoregressive modeling. The algorithm uses a piece-wise autoregressive (PAR) model to predict the unknown pixels of high resolution image. For this purpose, we used a block-based prediction model to predict the unknown pixels. The unknown pixels are categorized into three categories and they are predicted using predictors of different structure and order. Prediction accuracy and the visual quality of the interpolated image depend on the size of the window. We experimentally found an appropriate window size and have shown that subjective as well as objective (PSNR) quality of the high resolution (HR) images is same, on an average, as that of the competitive such method reported in literature and also the method is a single pass.

Retinal color images play an important role in supporting medical diagnosis. Digital retinal image usually are represented in such a large data volume that takes a considerable amount of time to be accessed and displayed from remote site. This paper aims to conduct a color retinal image coding using Entropy-Constrained Vector Quantization (ECVQ). In this paper, we use two objective parameters: Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR). Coded image which has the best quality of subjective and objective is the image coded with the value of λ = 0.1 and rate = 4.5 bpp.

In the problem of determining number of clustering and initial cluster centers, the mountain clustering algorithm was a simple and effective algorithm, it was a kind of clustering algorithm which could cluster sample set approximately and also could be used as the basis of other cluster analysis, which could provide initial cluster centers for other clustering algorithms. The improved algorithm of it was subtractive clustering, which had a great improvement in solving the problem of low efficiency of large sample set for mountain clustering, but its adaptability was not perfect. Therefore, put forward the regionalism adaptable mountain clustering algorithm, which based on the traditional mountain clustering algorithm divided sample set into regions and chose sample points of the largest weight to calculate their best initial value. Experimental results showed that the algorithm had stronger adaptability and accuracy of clustering, moreover speed was improved.

A more efficient newer algorithm of detecting systolic and diastolic pressure of human body along with a complete package of an effective user-friendly embedded programmable blood pressure monitoring system has been proposed in this paper to reduce the overall workload of medical personals as well as to monitor patient's condition more conveniently and accurately. Available devices for measuring blood pressure have some problems and limitations in case of both analog and digital devices. The sphygmomanometer, being analog device, is still being used widely because of its reliability and accuracy over digital ones. But it requires a skilled person to measure the blood pressure and obviously not being automated as well as time consuming. Our proposed system being a microcontroller based embedded system has the advantages of the available digital blood pressure machines along with a much improved form and has higher accuracy at the same time. This system can also be interfaced with computer through serial port/USB to publish the measured blood pressure data on the LAN or internet. The device can be programmed to determine the patient's blood pressure after each certain interval of time in a graphical form. To sense the pressure of human body, a pressure to voltage transducer is used along with a cuff in our system. During the blood pressure measurement cycle, the output voltage of the transducer is taken by the built-in ADC of microcontroller after an amplifier stage. The recorded data are then processed and analyzed using the effective software routine to determine the blood pressure of the person under test. Our proposed system is thus expected to certainly enhance the existing blood pressure monitoring system by providing accuracy, time efficiency, user-friendliness and at last but not the least the 'better way of monitoring patient's blood pressure under critical care' all together at the same time.

Road signs play an important role in our daily life which used to guide drivers to notice variety of road conditions and cautions. They provide important visual information that can help drivers operating their vehicles in a manner for enhancing traffic safety. The occurrence of some accidents can be reduced by using automatic road signs recognition system which can alert the drivers. This research attempts to develop a warning system to alert the drivers to notice the important road signs early enough to refrain road accidents from happening. For solving this, a non-linear weighted color enhancement method by pixels is presented. Due to the advantage of proposed method, different road signs can be detected from videos effectively. With suitably coefficients and operations, the experimental results have proved that the proposed method is robust, accurate and powerful in road signs detection.

The purpose of this study was to propose the method to recognize gestures based on neural networks and inertia sensor which recognizes the motions of the user using inertia sensor and lets the user enjoy the game by comparing the recognized gestures with the pre-defined gestures for the dance game contents.

A level set method based on the Bayesian risk and estimation of prior probabilities is proposed for image segmentation. First, the Bayesian risk is formed by false-positive and false-negative fraction in a hypothesis test. Second, through minimizing the average risk of decision in favor of the hypotheses, the level set evolution functional is deduced for finding the boundaries of targets. Third, the concave property of Kullback-Leibler information number is used to estimate the prior probabilities of each phase. Fourth, to prevent the propagating curves from generating excessively irregular shapes and lots of small regions, curvature and gradient of edges in the image are integrated into the functional. Finally, the Euler-Lagrange formula is used to find the iterative level set equation from the derived functional. Compared with other level-set methods, the proposed approach relies on the optimum decision; thus the approach has more reliability in theory and practice. Experiments show that the proposed approach can accurately extract the complicated textured and medical images; moreover, the algorithm is extendable for multiphase segmentation.

Edge detection is a basic and important subject in computer vision and image processing. An edge detector is defined as a mathematical operator of small spatial extent that responds in some way to these discontinuities, usually classifying every image pixel as either belonging to an edge or not. Many researchers have been spent attempting to develop effective edge detection algorithms. Despite this extensive research, the task of finding the edges that correspond to true physical boundaries remains a difficult problem.Edge detection algorithms based on the application of human knowledge show their flexibility and suggest that the use of human knowledge is a reasonable alternative. In this paper we propose a fuzzy inference system with two inputs: gradient and wavelet details. First input is calculated by Sobel operator and the second is calculated by wavelet transform of input image and then reconstruction of image only with details subimages by inverse wavelet transform. There are many fuzzy edge detection methods, but none of them utilize wavelet transform as it is used in this paper. For evaluating our method, we detect edges of images with different brightness characteristics and compare results with canny edge detector. The results show the high performance of our method in finding true edges.

Human faces contain a lot of demographic information such as identity, gender, age, race and emotion. Human being can perceive these pieces of information and use it as an important clue in social interaction with other people. Race perception is considered the most delicacy and sensitive parts of face perception. There are many research concerning image-base race recognition, but most of them are focus on major race group such as Caucasoid, Negroid and Mongoloid. This paper focuses on how people classify race of the racial closely related group. As a sample of racial closely related group, we choose Japanese and Thai face to represents difference between Northern and Southern Mongoloid. Three psychological experiment was performed to study the strategies of face perception on race classification. As a result of psychological experiment, it can be suggested that race perception is an ability that can be learn. Eyes and eyebrows are the most attention point and eyes is a significant factor in race perception. The Principal Component Analysis (PCA) was performed to extract facial features of sample race group. Extracted race features of texture and shape were used to synthesize faces. As the result, it can be suggested that racial feature is rely on detailed texture rather than shape feature. This research is a indispensable important fundamental research on the race perception which are essential in the establishment of human-like race recognition system.

Images are often significantly degraded by fog and their values are greatly reduced. According to retinex theory and the properties of fogged images, fog degradation can be eliminated by modifying illumination variation. The multi-scale retinex (MSR) algorithm was analyzed and an enhancement method directing to fogged images was proposed. Firstly, a preliminary global luminance was adjusted by linear stretching and screen algorithm to increase the luminance of darker pixels and compress the dynamic range as well. Then the local contrast was increased by an improved algorithm based on MSR. Finally, the output of local contrast enhancement was processed by the normal clipping stretching to realize color correction. Experiments show that the algorithm can effectively remove fog degradation from color images.

In this paper we are describing some important state-of the-art algorithms used for Image interpolation.These algorithms are broadly classified as prediction based and transform based methods. Motivation behind this work is to provide new researchers a detailed analysis of such algorithms in the context of artifacts, subjective and objective quality of interpolated image, computational cost and to give future research direction based on the analysis. However, the goal of this study was not to determine an overall best method, but to present a comprehensive catalogue of methods in a uniform terminology, to define general properties and requirements local techniques, and to enable the reader to select that method which is optimal for his specific application.

A self teaching image processing and voice recognition based system is developed to educate visually impaired children, chiefly in their primary education. System comprises of a computer, a vision camera, an ear speaker and a microphone. Camera, attached with the computer system is mounted on the ceiling opposite (on the required angle) to the desk on which the book is placed. Sample images and voices in the form of instructions and commands of English, Urdu alphabets, Numeric Digits, Operators and Shapes are already stored in the database. A blind child first reads the embossed character (object) with the help of fingers than he speaks the answer, name of the character, shape etc into the microphone. With the voice command of a blind child received by the microphone, image is taken by the camera which is processed by MATLAB® program developed with the help of Image Acquisition and Image processing toolbox and generates a response or required set of instructions to child via ear speaker, resulting in self education of a visually impaired child. Speech recognition program is also developed in MATLAB® with the help of Data Acquisition and Signal Processing toolbox which records and process the command of the blind child.

In this paper, we compare two correlation techniques for dense matching used in image corresponding problem, namely, the Sum of Squared Difference (SSD) and Normalized Cross Correlation (NCC). Both algorithms look for part of the image that matches a template based on intensity information. The window of the template is of Voronoi size, according to each Voronoi cells. The corresponding seed relations in each cell until all pixels within each cell are processed using SSD and NCC algorithms. In our experiments compare the performance of SSD and NCC in image with additive Gaussian noise, salt and pepper noise, and speckle noise. We found that SSD is more robust to noise than NCC in all cases.

This paper presents an improved algorithm for the image motion restoration by combining Wiener filtering with image histogram Equalization. We take the following steps to recover a uniform rectilinear motion image blur: Firstly, the parameter in the process of image degradation is determined on its spectrograph (Point Spread Function, PSF); Secondly, putting the PSF back into the Wiener filter's formula is to design the appropriate parameter y; Finally, the Wiener filtering method and the histogram equalization are integrated to form an improved Wiener filter algorithm for restoring uniform motion image blur. Experimental results show that visual comparison of images through experiments, improved the recovery algorithm to be significantly better than average effect of Wiener filtering method.

Computer-aided diagnosis (CAD) has become one of the major research subjects in medical imaging and diagnostic radiology. With cad, radiologists use the computer output as a "second opinion" and make the final decisions. Retrieving images is a useful tool to help radiologist to check medical image and diagnosis. The impact of contentbased access to medical images is frequently reported but existing systems are designed for only a particular context of diagnosis. The challenge in medical informatics is to develop tools for analyzing the content of medical images and to represent them in a way that can be efficiently searched and compared by the physicians. CAD is a concept established by taking into account equally the roles of physicians and computers. To build a successful computer aided diagnostic system, all the relevant technologies, especially retrieval need to be integrated in such a manner that should provide effective and efficient pre-diagnosed cases with proven pathology for the current case at the right time. In this paper, it is suggested that integration of content-based image retrieval (CBIR) in cad can bring enormous results in medicine especially in diagnosis. This approach is also compared with other approaches by highlighting its advantages over those approaches.

The segmentation of images into meaningful and homogenous regions is a key method for image analysis within applications such as content based retrieval. The watershed transform is a well-established tool for the segmentation of images. However, watershed segmentation is often not effective for textured image regions that are perceptually homogeneous. In order to properly segment such regions the concept of the "texture gradient" is now introduced. Texture information and its gradient are extracted using a combination of complex and packet wavelet transform. A novel marker and scan based watershed algorithm is then used to properly segment the identified regions. The combined algorithm produces effective texture and intensity based segmentation for the application to content-based image retrieval.

Unlike user created password, number of biometrics is limited for creating account in different organizations. Transformed biometrics attempts to solve the problem by transforming the biometric into another form, which is unique to the particular organization. This makes the availability of different transformed biometrics in different organizations transformed from the same biometrics and helps in foolproof transactions. In this article a novel approach to transformed biometrics using successive projection technique is suggested .In the proposed technique, the user can register up to 5*4^n-1 organizations if the length of the biometric password is 'n'.

The process of desertification, which extends from a long time ago, became a reality in Brazil. This phenomenon can be understood as land degradation, caused by factors including climatic changes and human activities. Besides being a threat to biodiversity, causes loss of soil productivity, threatening the lives of thousands of people living in affected regions. So, the identification of affected areas is essential to diagnose and prevent the problem. Satellite image has been a source of relatively low cost and widely used in this task. Therefore, is proposed in this study, a method to extract automatically areas heavily affected by desertification. The method is based on concepts of mathematical morphology, vegetation index and classification of digital images. Experiments are conducted separately, with images of CBERS 2 and 2B, and subsequently compared. The validation is done by crossing the results obtained with a reference image, created by a manual process.

Because of the complex nature of cells, the ability to distinguish a cell from the background of an image for automatic quantification remains a challenging task. Here, we describe a new technique for cell segmentation using an extended h-maxima transformation to find possible cell locations and a watershed algorithm. A novel method that is adopted to track the cells between image sequences is also discussed in this paper. The over segmentation problem of watershed algorithm is reduced by morphologic erosion, allowing for more accurate quantification, even in low contrast images. The number of cells and the average cell size could be determined in the image. Application of this method to a difficult dataset allowed us to identify 96% of the cells in the image and showed promising results for tracking cells between consecutive images.

A wavelet-based image watermarking scheme is proposed, based on insertion of 'logo' image as watermark in midfrequency domain. This new approach provides flexibility in determining the pixel to be watermarked and increases the data hiding capacity. It is easy to implement watermark embedding algorithm as well as the corresponding detection algorithm. The watermarking algorithm is tested under different attacks such as median filtering, image cropping and image compression. It is also robust. The experimental results prove that the method is more tamper proof and less perceptible for any type of images other than well known private methods in frequency domain. In the proposed approach, an original image is decomposed into wavelet coefficients then watermark is embedded through algorithm. The wavelet transform filters can be used as security key for the extraction of inserted watermark. The proposed watermark extraction technique is independent of the original image. The watermark embedded image is produced by taking the inverse 2-D discrete wavelet transform of the altered wavelet decomposition. Here we have given the relation between the area of the channel in which we insert the watermark and the area affected in original image.

The paper provides a pre-stage of any data hiding method hiding message data in media data and a data extraction method of extracting the hidden data, wherein message data is dispersively hidden in digital media data, such as images, to prevent a third person from forging/modifying the message data easily. More specifically, the technique relates to a data hiding method in which media data is expressed as a media array while message data is expressed as a message array so that the array elements of the message array can be dispersively hidden in the media array randomly by scrambling order of particular array element of the media array based on a private key. It needs to declare that the proposed strategy aims only to enhance the watermark security. It is not used to improve the robustness of watermark. The current version of this paper has had a correction made to it at the request of the author. Please see the linked Errata for further details.

Image segmentation is the process of dividing an image into homogenous regions. It is an essential step towards high-level image processing task such as image analysis, pattern recognition and computer vision. Processing of color images has become an important issue due to its huge usage in computer vision applications. It is observed that most of the color image segmentation techniques are derived from monochrome image segmentation. The techniques for segmentation of monochrome images are based on the principles of histogram thresholding, edge detection, region growing etc. Many color image segmentation algorithms using different color models and these principles are proposed. Extraction of objects within an image without a prior knowledge is one of the important issues in segmentation area. Novel approaches such as fuzzy set theory, neural network and neuro-fuzzy based segmentation are coming up to tackle this problem. This paper is an endeavor to review various algorithms and recent advances in color image segmentation.

The objective of the 3D shape estimation from focus is to estimate depth map of the scene or object based on best focus points from camera lens. In shape from focus (SFF), the measure of focus - sharpness - is the crucial part for final 3D shape estimation. However the noise imposed during image acquisition process by imaging system prevents exact focus measure. The traditional noise filters remove not only noise but also sharpness information. In this paper, mean shift algorithm was applied to remove noise imposed by the imaging process while minimizing loss of informative edges. Experimental results show that the mean shift algorithm can be applied before computing focus measure from image sequence corrupted by Gaussian noise and Impulse noise. Applying mean shift filtering before computing focus measure is promising in case the noise type during image acquisition is not known.

This paper attempts to discuss the evolution of semantic based image retrieval. The explosive growth of image data leads to the need of research and development of Image retrieval. Image retrieval researches are moving from keyword, to low level features and to semantic features. Drive towards semantic features is due to the problem of the keywords which can be very subjective and time consuming while low level features cannot always describe high level concepts in the users' mind. Framework of semantic based image retrieval as well as the processes involved has been discussed. This paper also highlights both the already addressed and outstanding issues.

Ordered dither is one of the popular techniques for digital halftoning where the original continuous tone image is thresholded against an orderly generated screen matrix. This paper presents a technique to generate the screen matrix using three-layer back-propagation multi layer perceptron (BP-MLP) artificial neural network (ANN) model. The image raw data has been preprocessed prior feeding to the input layer. The output obtained at the hidden layer of the model has been considered as the screen matrix for ordered dither. The results achieved using this technique have been evaluated subjectively as well as objectively using commonly used quality indices like peak signal to noise ratio (PSNR), universal quality index (UQI) and structural similarity index measure (SSIM).

Iris is one of the most discriminating human physiological traits being used for personal human identification. The success of iris based system is highly relied on the accurately captured and precisely segmented iris image. False rejection rate has been a major challenge in the success of such system which primarily results from inaccurate iris segmentation. Most of the presented algorithms on iris segmentation considers pupil as perfect circle. However, according to observation this is not true in all the cases. In addition, a little angular shift in position of subject iris can further deteriorate the performance of algorithms based on circular assumption. To improve the quality of segmentation, an effective algorithm is proposed for iris segmentation which takes into consideration issues related to irregular pupil boundary as well as computation intensive nature of the prevailing algorithms. Contrary to all the previous approaches, the proposed algorithm is based on detection of pupil prospect point within the pupil region and utilizes bi-directional horizontal projections and distance parameters to detect the pupillary as well as limbus boundaries. The processing involved is linear in nature. Simulation of the proposed algorithm is done in Matlab.

In this a paper, a feature based approach to face detection has been proposed using an ensemble of algorithms. The method uses chrominance values and edge features to classify the image as skin and nonskin regions. The edge detector used for this purpose is Laplacian of Gaussian (LoG) which is found to be appropriate when images having multiple faces with noise in them. Eight connectivity analysis of these regions will segregate them as probable face or nonface. The procedure is made more robust by identifying local features within these skin regions which include number of holes, percentage of skin and the golden ratio. The method proposed has been tested on color face images of various races obtained from different sources and its performance is found to be encouraging as the color segmentation cleans up almost all the complex facial features. The result obtained has a calculated accuracy of 86.5% on a test set of 230 images.

Works on Facial Expression Recognition (FER) have mostly been done using image based approaches. However, in recent years, researchers have also been trying to explore the use of 3D information for the task of FER. Most of the time, there is a need for having a neutral (expressionless) face of the subject in both the image based and 3D model based approaches. However, this might not be practical in many applications. This paper tries to address this limitations in previous works by proposing a novel technique of feature extraction which does not require any neutral face of the subjects. It has been proposed and validated experimentally that the motion of some landmark points on the face, in exhibiting a particular facial expression, is similar in different persons. Separate classifier is made and relevant feature points are selected for each expression. One vs all SVM classification gives promising results.

Forensic image analysis is required an expertise to interpret the content of an image or the image itself in legal matters. Major sub-disciplines of forensic image analysis with law enforcement applications include photo-grammetry, photographic comparison, content analysis and image authentication. It has wide applications in forensic science range from documenting crime scenes to enhancing faint or indistinct patterns such as partial fingerprints. The process of forensic image analysis can involve several different tasks, regardless of the type of image analysis performed. Through this paper authors have tried to explain these tasks, which are described in to three categories: Image Compression, Image Enhancement & Restoration and Measurement Extraction. With the help of examples like signature comparison, counterfeit currency comparison and foot-wear sole impression using the software Canvas and Corel Draw.

A new image fusion method based on Multiscale Geometric Analysis (MGA), which uses the improved fusion rules, is put forward in this paper. Firstly, the input low-level-light image and infrared image are decomposed by Curvelet transform, which is realized by Unequally-Spaced Fast Fourier Transforms. Secondly, the decomposed coefficients in different scales and directions are fused by corresponding fusion rules. At last, the fusion image is acquired by recomposing the fused coefficients. The simulation results show that this method performs better than the conventional wavelet method both in the subjective vision aspect and the objective estimation indices.

Traditional method of pattern recognition confuse difference tow procession of pattern memory (microcosmic layer) and pattern classifying (macroscopic layer), it is main cause which the pattern methods are difficulty applied to solve really problem. A new frame of APRF (Agent-based Pattern Recognition Frame) is proposed based on Agent theory. The study goal of APRF are cognizing pattern from integral view, erecting the bridge between microcosmic layer and macroscopic layer and uncovering perplex of pattern emerge.

Large collection of information is being created in many areas of modern life on daily basis. This information exists in many forms from plain text to high resolution multimedia. Today computers are many times faster than human in text based searching using keywords and indexing but the story is totally different in case of multimedia. In context of image retrieval, acquiring storing, sorting and transmitting photos is now trivial, but it is significantly harder to manipulate, index, sort, filter or search through them. The research presents an overview of different techniques used in contentbased image retrieval (CBIR) systems and what are some of the proposed ways of querying such searches that are useful when specific keywords for the object are not known. Advances, applications and problems in content-based image retrieval are also discussed. Moreover a system is also developed for content based image retrieval and tested with two database containing 1,000 and 10,000 images respectively.

The accumulation of large collections of digital images has created the need for efficient and intelligent schemes for content-based image retrieval. Our goal is to organize the contents semantically, according to meaningful categories. Automatic annotation is the process of automatically assigning descriptions to an image or video that describes the contents of the image or video. In this paper, we examine the problem of automatic captioning of multimedia containing round and square objects. On a given set of images and videos we were able to recognize round and square objects in the images with accuracy up to 80% and videos with accuracy up to 70%.

Signature verification is one of the most widely researched areas in document analysis and signature biometric. Various methodologies have been proposed in this area for accurate signature verification and forgery detection. In this paper we propose a unique two stage model of detecting skilled forgery in the signature by combining two feature types namely Sum graph and HMM model for signature generation and classify them with knowledge based classifier and probability neural network. We proposed a unique technique of using HMM as feature rather than a classifier as being widely proposed by most of the authors in signature recognition. Results show a higher false rejection than false acceptance rate. The system detects forgeries with an accuracy of 80% and can detect the signatures with 91% accuracy. The two stage model can be used in realistic signature biometric applications like the banking applications where there is a need to detect the authenticity of the signature before processing documents like checks.

Data mining refers to the process of extracting knowledge that is of interest to the user. Traditional data mining techniques have been developed mainly for structured data types. The image data type does not belong to this structured category, suitable for interpretation by a machine and hence the mining of image data is a challenging problem. Accordingly, in image mining, an image retrieval system is a computer system that can browse, search and retrieve images from a large database of digital images. This research work is aimed at compression and retrieval of images from large image archives. A Kohonen Self Organization Map approach using content categorization, including feature level clustering, is developed to provide a differential compression scheme. It ensures that the visual features are mapped to codebooks, which significantly speed up content-based retrieval. The interaction between compression and content indexing are proposed, which include techniques for feature extraction, indexing, and categorization. K-means clustering algorithm is used to build the feature cluster. This approach leads to the similarity matching based on shape and texture, which supports functions like "query by example". Experimental results demonstrate that the proposed method can improve the compression ratio compared to VQ. The average retrieval time is less than 2seconds, which is proved to be efficient.

There has been considerable interest in immersive and realistic virtual environment system and how to improve human and computer interaction has been a main challenge. For vision-based human-computer interaction, extraction of user region from camera images is an essential part. In this paper, we propose a background subtraction method that segments dynamic projected background in a rear-projection-based virtual environment system. In the projector-camera system, the projected background is inherently known by the projector input images although its appearance is changed by the geometric and radiometric transformation between projector and camera. Therefore, we can compute the expected background location and appearance based on geometric and radiometric calibration of projector-camera system and thus separate user region from dynamic projected background by simple subtraction between camera images and the computed background. Experimental results are given for verifying the usefulness of the proposed method.

Plum Blossom as the Chinese traditional flowers may be unique all over the world and has the first right of access to international registry of flower. In this paper, the SpeedTree software is used to quickly build reality model of the plum tree. The graphics texture mapping techniques is used, and the plum tree image maps express the geometric model of the surface material, which constitutes a visual image of the graphic objects. It is significant for non-destructive study of plum and virtual garden.

In the biometric recognition, face recognition is the most natural, direct method. Research on face recognition has a high theoretical significance and practical value. In this paper, firstly we use the Gabor filter to extract face image features, and then denote to further dimensionality reduction by Hopfield Neural Network. At last, for face classification, a new method based on support vector machine- No-balance Binary Tree Support Vector Machine (NBBTSVM) is proposed to decide a label in this face recognition task. SVM has excellent performance to solve binary classification but for multi-classification, it's an ongoing research. According to our experiment results, NBBTSVM could do a good performance.

Now-a-days Multimedia data can be easily replicated and the copyright is not legally protected. Cryptography does not allow the use of digital data in its original form and once the data is decrypted, it is no longer protected. Here we have proposed a new double protected digital image watermarking algorithm, which can embed the watermark image blocks into the adjacent regions of the host image itself based on their blocks similarity coefficient which is robust to various noise effects like Poisson noise, Gaussian noise, Random noise and thereby provide double security from various noises and hackers. As instrumentation application requires a much accurate data, the watermark image which is to be extracted back from the watermarked image must be immune to various noise effects. Our results provide better extracted image compared to the present/existing techniques and in addition we have done resizing the same for various displays. Adaptive resizing for various size displays is being experimented wherein we crop the required information in a frame, zoom it for a large display or resize for a small display using a threshold value and in either cases background is not given much importance but it is only the fore-sight object which gains importance which will surely be helpful in performing surgeries.

This paper proposes an edge detection method by Canny operator with a feedback strategy for glomeruli extraction. As we know, the effect of the Canny operator is determined by three parameters: high threshold, low threshold and standard deviation. To obtain the appropriate parameters for each image, Otsu method is used to set high and low thresholds of Canny operator firstly. And then, to select the optimal standard deviation, a feedback strategy is developed. After parameter selection, Canny operator is applied to our renal biopsy images and experimental results show that some samples can achieve successful extraction and the others result in the discontinuous edges of glomeruli. In the case of the latter, the endpoints in an image should be located and connected to form a whole edge of glomerulus. The experiments have produced the promising results for our samples.

Active learning with support vector machine(SVM) selects most informative unlabeled images for user labeling, however small training samples affect its performance. To improve active learning and use more unlabeled data, we propose a new algorithm training three SVMs separately on the color, texture and shape features of labeled images with three different kernel functions according to the features' distinct statistical properties. Different algorithms are used in the selection of disagreement and agreement samples from unlabeled data and also in the calculation of their confidence degrees. The lowest confident disagreement samples are returned to user to label and added to the training data set with the highest confident agreement samples. Experimental results verify the high effectiveness of our method in image retrieval.

The growing capacity of computers, the abundance of digital cameras and the increased connectivity of the world all point to large digital multimedia archives. They include images and videos from the World Wide Web, museum objects, flowers, trademarks, and views from everyday life. The faster they grow, the more prominently needed is the efficient access to the content of the images and videos. In this paper we have given important step of feature extraction, will be discussed in detail such as color, shape and texture information, particularly paying attention to discriminatory power and invariance. Then, we focus on the concepts of indexing and genre classification as intermediate step to sort the data. We pay attention to (interactive) ways to perform browsing and retrieval by means of information visualization and relevance feedback. Methods are being discussed to localize the retrieved objects in images. We adopt a hybrid approach for such text extraction by exploiting a number of characteristics of text blocks in color images and video frames. Our system detects both caption text as well as scene text of different font, size, color and intensity. Such texts are used for retrieval of video clips based on any given keyword. Content-Based Image And Video Retrieval addresses the basic concepts and techniques for designing content-based image and video retrieval systems.

In the fields of image processing, signal processing and recognition, image Segmentation is an efficient method for segmenting the phonocardiograph signals (PCG) is offered. Primarily, inter-beat segmentation is approved and carried out by means of DII lead of the ECG recording for identifying the happenings of the very first heart sound (S1). Then, the intra-beat segmentation is attained by the use of recurrence time statistics (RTS), and that is very sensitive to variations of the renovated attractor in a state space derived from nonlinear dynamic analysis. Apart from this if the segmentation with RTS is unsuccessful, a special segmentation is proposed using threshold that is extracted from the high frequency rate decomposition and the feature extraction of the disorder is classified based on the murmur sounds. In the Inter-beat segmentation process the accuracy was 100% of the over all PCG recording. Taking into account a different level of PCG beats were strongly concerned by different types of cardiac murmurs and intra-beat segmentation are give up for an accurate result.

Human recognition based on biometrics finds many important applications in many life sectors and in particular in commercial and law enforcement. This paper aims to give a general overview of the advances in the biometric recognition methods. We concentrate on main methods and accessible ideas presented for human recognition systems based on two types of biometrics: iris and fingerprint. We present a quick overview of the landmark papers that laid the foundation in each track then we present the latest updates and important turns and solutions that developed in each track in the last few years.

In this paper we introduce a new approach for information retrieval from Persian document image database without using Optical Character Recognition (OCR).At first an attribute called subword upper contour label is defined then, a pictorial dictionary is constructed based on this attribute for the subwords. By this approach we address two issues in document image retrieval: keyword spotting and retrieval according to the document similarities. The proposed methods have been evaluated on a Persian document image database. The results have proved the ability of this approach in document image information retrieval.

The ISO-945² committee has defined six classes of grain morphology through reference drawings for cast iron graphite grain classification. These reference drawings are universally accepted for classification of graphite grains. The main aim of this work is to propose a neural network approach for cast iron classification based on graphite grain morphology by processing microstructure images. The two sets of shape features investigated are, Simple Shape Descriptors (SSDs) and Moment Invariants(MIs). The classifiers like, feed forward neural network with back propagation and radial basis functions are also investigated. The experimentation is carried out using the metallographic images from the well known microstructures library⁴. For training and testing the networks, the grain shapes identified in ISO-945 reference drawings and the grain classification by the experts are used. The moment invariant shape features and neural network classifier with radial basis function yield better classification results for graphite grains.

It is well known that image processing depends heavily upon image representation technique . This paper intends to find out the optimal path of mobile robots for a specified area where obstacles are predefined as well as modified. Here the optimal path is represented by using the Quad tree method. Since there has been rising interest in the use of quad tree, we have tried to use the successive subdivision of images into quadrants from which the quad tree is developed. In the quad tree, obstacles-free area and the partial filled area are represented with different notations. After development of quad tree the algorithm is used to find the optimal path by employing neighbor finding technique, with a view to move the robot from the source to destination. The algorithm, here , permeates through the entire tree, and tries to locate the common ancestor for computation. The computation and the algorithm, aim at easing the ability of the robot to trace the optimal path with the help of adjacencies between the neighboring nodes as well as determining such adjacencies in the horizontal, vertical and diagonal directions. In this paper efforts have been made to determine the movement of the adjacent block in the quad tree and to detect the transition between the blocks equal size and finally generate the result.

In the engineering project, arbitrary waveform generator controlled by software interface is needed by simulation and test. This article discussed the program using the SCPI (Standard Commands For Programmable Instruments) protocol and the VISA (Virtual Instrument System Architecture) library to control the Agilent signal generator (Agilent N5182A) by instrument communication over the LAN interface. The program can conduct several signal generations such as CW (continuous wave), AM (amplitude modulation), FM (frequency modulation), ΦM (phase modulation), Sweep. As the result, the program system has good operability and portability.

In order to meet the requirement of obtaining the object figure quickly and accurately, XJTUDP software has been developed successfully by oneself based on photogrammetry theory. The contents of composing and explored of this system are introduced in this paper. The VDI/VDE2634 testing program is taken as referenced project, the self-designed framework of cube is taken as tested target, and then the conclusion that precision of XJTUDP may meet the standard of industrial measurement has been reached out. Finally, large-scale waterwheel leave is taken as example to conduct measuring, it is proved that photogrammetry system explored by ourselves may be imposed applying on the field of industrial measuring successfully.

This paper presents a technique for texture feature extraction and classification using wavelet transform. A image is decomposed into no. of sub-bands after applying Wavelet transform to it. A three level decomposition is carried out. A number of sub-bands are generated after wavelet decomposition. An energy signature is computed for each sub-band of these wavelet coefficients. A k-nearest neighbor's classifier is then employed to classify texture patterns. To test and evaluate the method, several sets of textures along with different wavelet bases are employed. Experimental results show superiority of the proposed method.

In retinal image analysis, the detection of optic disk is of paramount importance. It facilitates the tracking of various anatomical features and also in the extraction of exudates, drusens etc., present in the retina of human eye. The health of retina crumbles with age in some people during the presence of exudates causing Diabetic Retinopathy. The existence of exudates increases the risk for age related macular Degeneration (AMRD) and it is the leading cause for blindness in people above the age of 50.A prompt diagnosis when the disease is at the early stage can help to prevent irreversible damages to the diabetic eye. Screening to detect diabetic retinopathy helps to prevent the visual loss. The optic disk detection is the rudimentary requirement for the screening. In this paper few methods for optic disk detection were compared which uses both the properties of optic disk and model based approaches. They are uniquely used to give accurate results in the retinal images.

Tuberculosis (TB) is a disease caused by bacteria called Mycobacterium tuberculosis. It usually spreads through the air and attacks low immune bodies such as patients with Human Immunodeficiency Virus (HIV). This work focuses on finding close association rules, a promising technique in Data Mining, within TB data. The proposed method first normalizes of raw data from medical records which includes categorical, nominal and continuous attributes and then determines Association Rules from the normalized data with different support and confidence. Association rules are applied on a real data set containing medical records of patients with TB obtained from a state hospital. The rules determined describes close association between one symptom to another; as an example, likelihood that an occurrence of sputum is closely associated with blood cough and HIV.

In diagnosing lung diseases using x-ray images of a human thorax, there is a huge risk of error in detecting abnormalities of the lung. This may be caused by geometric differences in the images that are being compared. To minimize the possible errors, a system is proposed to assist in the diagnosis process. In implementing this system, a registration process of the images is required as the first step in minimizing the human errors. A feature based method is used to solve the registration of images by using a scale invariant feature transform (SIFT) as the method of feature extraction. Using this feature based method is hoped to result in a better registration than the area based method that was previously used.

Here a fuzzy expert system design for diagnosing, analyzing and learning purpose of the cancer diseases is described. For this process prostate specific antigen (PSA), age and prostate volume (PV) has been used as an input parameters and prostate cancer risk (PCR) as an output. This system allows determining if there is a need for the biopsy and it gives to user a range of the risk of the cancer diseases. It is observed that this system is rapid, economical, without risk than traditional diagnostic systems, has also a high reliability and can be used as learning system for medicine students.

The blocking effect is a major drawback of the DCT-based compression scheme at low bit rates. Significantly decreasing blocking effects can raise compression ratios for a particular image quality, or improve the quality with regard to the specific bit rate of compression. This work presents a scheme based on two mode filters in terms of the activity across block boundaries. For smooth regions, the strong smooth filter exploits the correlation between the neighboring blocks to reduce the discontinuity of the pixels across the boundaries. The weak smooth filter employs an edge-preserving smooth filter for texture and edge-based regions. Simulation results reveal that the proposed algorithm significantly lowers the blocking artifact, as judged by both objective and subjective measures.

A method of detecting multiple barcodes simultaneously using time division multiplexing technique has been proposed in this paper to minimize the effective time needed for handling multiple tags of barcodes and to lessen the overall workload. Available barcode detection systems can handle multiple types of barcode but a single barcode at a time. This is not so efficient and can create large queue and chaos in places like mega shopping malls or large warehouses where we need to scan huge number of barcodes daily. Our proposed system is expected to improve the real time identification of goods or products on production lines and in automated warehouses or in mega shopping malls in a much more convenient and efficient way. For identifying of multiple barcodes simultaneously, a particular arrangement and orientation of LASER scanner and reflector have been used with a special curve shaped basement where the barcodes are placed. An effective and novel algorithm is developed to extract information from multiple barcodes which introduces starting pattern and ending pattern in barcodes with bit stuffing technique for the convenience of multiple detections. CRC technique is also used for trustworthiness of detection. The overall system enhances the existing single barcode detection system by a great amount although it is easy to implement and use.

In this paper we propose a system for classification problem of handwritten text. The system is composed of preprocessing module, supervised learning module and recognition module on a very broad level. The preprocessing module digitizes the documents and extracts features (tangent values) for each character. The radial basis function network is used in the learning and recognition modules. The objective is to analyze and improve the performance of Multi Layer Perceptron (MLP) using RBF transfer functions over Logarithmic Sigmoid Function. The results of 35 experiments indicate that the Feed Forward MLP performs accurately and exhaustively with RBF. With the change in weight update mechanism and feature-drawn preprocessing module, the proposed system is competent with good recognition show.

In view of the growing use of automatic speech recognition in the modern society, we study various alternative representations of the speech signal that have the potential to contribute to the improvement of the recognition performance. In this paper wavelet based features using different wavelets are used for Hindi digits recognition. The recognition performance of these features has been compared with Linear Prediction Coefficients (LPC) and Perceptual Linear Prediction (PLP) features. All features have been tested using Hidden Markov Model (HMM) based classifier for speaker independent Hindi digits recognition. The recognition performance of PLP features is11.3% better than LPC features. The recognition performance with db10 features has shown a further improvement of 12.55% over PLP features. The recognition performance with db10 is best among all wavelet based features.

Image fusion plays an important role in improving high resolution remote sensing images, as many Earth observation satellites provide both high-resolution panchromatic and multispectral images. To date, many image fusion techniques have been developed. Existing traditional image fusion techniques such as the intensity-hue-saturation (IHS) transform, wavelet transform and principal components analysis(PCA) methods may not be optimal for fusing the new generation commercial high-resolution satellite images such as IKONOS and Quick Bird. However, the available algorithms can hardly meet a satisfactory fusion requirement for high resolution remote sensing images. Among the existing fusion algorithms, the IHS technique is the most widely used one technique. But the color distortion of this technique is often obvious, especially when high resolution multispectral images are fused with its panchromatic images. This study presents a new fusion approach that integrates both IHS and histogram match techniques to reduce the color distortion of high resolution remote sensing fusion results. Different high resolution remote sensing images have been fused with this new approach. The result proves that the concept of the proposed improved IHS is promising, and it does significantly improve the fusion quality compared to conventional IHS transform fusion techniques.

Image Processing always aims at extracting maximum information from an image. To achieve this we have to analyze the image completely along its periphery. But the parts of an image are hardly straight, they contain continuously varying slopes. Wavelet based image processing gives low resolution when the image has largely varying slopes and they give redundant coefficients. If we tile the whole image, we get curve-lets meaning 'small curves'. If this tilling is optimum, we get parts of the curve which resemble to the straight lines. These straight lines are then analyzed and reconstructed using 'Curvelet Transform'. Curvelet Transform represents edges of a curve better than Wavelet Transform. This transform uses 'Ridgelet Transform' as a main processing. Ridgelet Transform is a two step process using Radon Transform and DWT. Radon transform analysis involves the mapping of rectangular coordinates into the polar or angular coordinates. With the increasing need for higher speed and lower memory requirement, we, in this paper propose to compute the Ridgelet coefficients without involving the conversion to angular coordinates. We have used Radon transforms our basic building block. As it will be seen taking 1-D DWT on Radon Transform results in Ridgelet Transform. At the end of the paper the images having many 'ridges', our transform gives better PSNR than Wavelet transform and many others. It also saves computational time by using fast FFT algorithm and avoiding operating on Tiles having less variation of pixels. The PSNR also depends on the algorithm used to perform DWT.

Cooperative networks have been appreciated for their spatial diversity benefits in wireless communications. In this paper a novel scheme of cooperative networks depending on the number and locations of relays in the network. The effect of relay number and locations are investigated by considering energy optimization. First selects the optimal relay from a set of available relays and then uses this "optimal" relay for cooperation between the source and the destination. The simulation-based performance analysis confirms that the cooperative relaying scheme has an advantage of diversity gain thus improving the bit error ratio performance. The simulation results demonstrate that the proposed cooperative relay node selection algorithm can improve performances by achieving the cooperative gain.

The multimedia information, especially video and audio information, regarded as a common data stream, with the traditional encryption technology encrypted, ignoring the characteristics of multimedia data, has some limitations. On the other hand, chaotic sequences have several good properties including the ease of their generation, their sensitive dependence on their initial condition and so on. Therefore, this paper discussed image encryption arithmetic on the basis of chaotic sequences through dispersing the real number value chaotic sequences into symbol matrix and transformation matrix, and then encrypted the image. Preliminary results proved that the image encryption arithmetic based on chaotic sequences possesses the traits, namely fast speed for encryption speed, perfect results for encryption.

Removing Noise from the image is a challenging problem for the researchers. This paper proposes a two phase threshold based median filtering technique for salt and pepper impulse noise removal. It is implemented as a two pass algorithm: In the first pass corrupted pixels are perfectly detected using min-max strategy and an adaptive working window based on estimated noise density. Second phase is a threshold based filtering technique to correct the corrupted pixels by a valid median. Experimental results have shown that the proposed technique performs far more superior than many of the efficient median based filtering techniques reported in the literature in terms of Peak Signal (PSNR) and visual perception of the images corrupted by impulse noise even to the tune of seventy percent.

This design introduces a used car on the design of LED decorative light strip. This LED named Knight LED. In This system we use ATMEGA8 as the Master MCU Chip. Through the microcontroller to implement the wireless remote control receiver and the LED lights of different modes of switching, different brightness control. Also we use ULN2803 as the LED driver.

Firstly, infrared image is decomposited using stationary wavelet transform, it is proposed based on stationary wavelet transform with Interscale and Intrascale Dependencies for infrared image denoising. Then the minimum mean square-error estimation is applyed to estimated coefficient. The wavelet coefficients are revised using the correlations between coefficients at the same scale. The denoised image is obtained through inverse wavalet transform. The experimental results show the infrared image can be denoised better than the method neglecting the correlations between Intrascales and have a well SNR as well as the visual quality.

In order to study application of digital control technology on DC/DC converter, principle of increment mode PID control algorithm was analyzed in the paper. Then, a SCM named STC12C5410AD was introduced with its internal resources and characteristics. The PID control algorithm can be implemented easily based on it. The output of PID control was used to change the value of a variable that is 255 times than duty cycle, and this reduced the error of calculation. The valid of the presented algorithm was verified by an experiment for a BUCK DC/DC converter. The experimental results indicated that output voltage of the BUCK converter is stable with low ripple.

The vehicle dispatching system (VDS) is a kind of real-time management system for moving objects. It is developed to meet with the requirement of vehicle orientation and dispatching. The GPS, mobile communication, data communication, and computer and GIS techniques are integrated into this system. With these techniques, the VDS is described, and its functions are analyzed. Then the communication between mobile terminals and the server is designed based on Winsock and Java multithreads technique.

This paper shows the effect that selected attribute filters have on existing watermarks of an image. Seven transform domain watermarking algorithms and five attributes have been investigated. The attributes are volume, gray-level, power, area and vision. Apart from only one, all of the filters have been found not to affect the underlying watermarks.

Introducing biometrics into information systems may result in considerable benefits. Most of the researchers confirmed that the finger print is widely used than the iris or face and more over it is the primary choice for most privacy concerned applications. For finger prints applications, choosing proper sensor is at risk. The proposed work deals about, how the image quality can be improved by introducing image fusion technique at sensor levels. The results of the images after introducing the decision rule based image fusion technique are evaluated and analyzed with its entropy levels and root mean square error.

This paper introduces a traffic light design solution using HB LED color mixture,namely an implementation method using a set of traffic lights composed of red, green and blue HB LED to replace ordinary red, yellow and green lights. This scheme realizes HB LED color mixture lighting design on the basis of Cypress EZ-Color controller and codeless embedded design software PSoC Express.

Pixels of a grayscale image are classically associated with a single grayscale value. However, capturing grayscale images comes along with two kinds of uncertainty: numerical uncertainty (do we measure the actual value of the pixel or just an approximation?) and spatial uncertainty (does the measured pixel correspond to the actual spatial position or has it shifted?). Interval-valued fuzzy set theory provides a framework to model grayscale images of which the captured grayscale values are uncertain. This is realized by associating every pixel with a closed interval of possible grayscale values instead of with one single value. Based on this image model, a new corresponding morphological framework to process these images (e.g., using dilation and erosion) has been developed. In that way, we are not only able to model the uncertainty that is present during image capturing, but we are also able to process it such that the information regarding the uncertainty is never lost. In this paper, we study the interval-valued fuzzy morphological model based on the minimum-operator. Properties that are relevant in the context of image processing, as well as some interesting decomposition and construction properties, are discussed. This study gives an insight in the morphological model and will help researchers when they want to apply it in practice.

The Plant is one of the hot fields in the current virtual reality modeling research, and undoubtedly an important component of factors in the natural scenes. It is difficult to be drawn in terms of shape, so there have be a lot of methods in the research of three-dimensional modeling of plants. This article describes the simulation modeling of plants and related technology research and development situation, summarizes the main research problems, and discusses the future research and application development trends and prospects.

This paper presents a system for retrieving information from a domain specific document collection made up of data rich unnatural language text documents. Instead of conventional keyword based retrieval, our system makes use of domain ontology to retrieve the information from a collection of documents. The system addresses the problem of representing unnatural language text documents and constructing a classifier model that helps in the efficient retrieval of relevant information. Query to this system may be either the key phrases in terms of concepts or a domain specific unnatural language text document. The classifier used in this system can also be used to assign multiple labels to the previously unseen text document belonging to the same domain. An empirical evaluation of the system is conducted on the domain of text documents describing the classified matrimonial advertisements to determine its performance.

Image de-noising is usually required to be performed before further processing like segmentation, object recognition and texture analysis. This paper gives a novel method of combining the filtration process and the post-processing techniques. For this purpose , evolvable hardware is combined with Genetic Algorithm to offer potential cost efficiency together with the flexibility of an adaptive system, producing a high-speed non-linear adaptive median filter . The spatial domain filtering techniques discussed are various forms of median filtering from which the 'FITTEST' algorithm is chosen depending on user requirements . Adaptive median filter removes the impulse & salt and pepper noise efficiently while retaining the edges and other detailed features. Further , after filtration , an object in the image is represented on a scale of categories and recognition algorithms are used to find the most detailed category according to information extracted from the image. The categorization is based on the color and texture content of the image. Border detection is also done by classifying the pixels as homogeneous and heterogeneous by comparison with the neighboring pixels . These techniques can be effectively implemented to process aerial photographs and to detect oil spills in the sea and properties of the ocean useful for fishery and navigation.

A novel denoising method based on the nonsubsampled wavelet-based contourlet transform (N-WBCT) was proposed in this paper. It employs wavelet transform for multi-scale decomposition, and nonsubsampled directional filter banks (NSDFB) for directional decomposition. N-WBCT has not only the properties of multi-resolution and multi-direction, but also the property of translation invariance, which is useful in eliminating the Gibbs phenomenon. The experiment results showed that this algorithm can get higher PSNR and the better visual.

Segmentation of icosahedral virus density map from cryo-electron microscope (CryoEM) is a challenging task because virus structure is complex and density map is at low resolution. Fast marching method is widely used in segmentation, in which seed selection is essential for correct segmentation results. However, the selection of an appropriate seed is difficult. In this paper, we present the method of selecting the seed in fast marching algorithm by making use of the shape symmetry to improve the fast marching method for icosahedral virus segmentation. Based on the feature of icosahedron, we compute and get its symmetry axes inside the density map. With these symmetry axes, we specify the initial seeds with the local maxima value along symmetry axes. Further, the new data structures are presented, which can effectively reduce the memory cost when implement the fast marching algorithm. Experimental results show that the approach can obtain segmentation results of the density maps fast and accurately.

Image data is a combination of information and redundancies, the information is part of the data be protected because it contains the meaning and designation data. Meanwhile, the redundancies are part of data that can be reduced, compressed, or eliminated. Problems that arise are related to the nature of image data that spends a lot of memory. In this paper will compare 31 wavelet function by looking at its impact on PSNR, compression ratio, and bits per pixel (bpp) and the influence of decomposition level of PSNR and compression ratio. Based on testing performed, Haar wavelet has the advantage that is obtained PSNR is relatively higher compared with other wavelets. Compression ratio is relatively better than other types of wavelets. Bits per pixel is relatively better than other types of wavelet.

Image Classification is the process of assigning classes to the pixels in remote sensed images and important for GIS applications, since the classified image is much easier to incorporate than the original unclassified image. To resolve misclassification in traditional parametric classifier like Maximum Likelihood Classifier, the neural network classifier is implemented using back propagation algorithm. The extra spectral and spatial knowledge acquired from the ancillary information is required to improve the accuracy and remove the spectral confusion. To build knowledge base automatically, this paper explores a non-parametric decision tree classifier to extract knowledge from the spatial data in the form of classification rules. A new method is proposed using a data structure called Peano Count Tree (P-tree) for decision tree classification. The Peano Count Tree is a spatial data organization that provides a lossless compressed representation of a spatial data set and facilitates efficient classification than other data mining techniques. The accuracy is assessed using the parameters overall accuracy, User's accuracy and Producer's accuracy for image classification methods of Maximum Likelihood Classification, neural network classification using back propagation, Knowledge Base Classification, Post classification and P-tree Classifier. The results reveal that the knowledge extracted from decision tree classifier and P-tree data structure from proposed approach remove the problem of spectral confusion to a greater extent. It is ascertained that the P-tree classifier surpasses the other classification techniques.

Considering the difficulties for traditional methods in clustering analysis of spatial data, in this paper, a novel spatial data clustering method based on an improved evolutionary algorithm is proposed. It effectively solved the two main problems puzzling many researchers, i.e., 1) difficulty in coping with the local optimum, and 2) sensibility to the center selections of the initial clustering. Empirical evaluation of our method indicates that it has better performance, compared with the other methods in literature.

A novel voice morphing method is proposed to make the speech of the source speaker sound like the voice uttered by a target speaker. This method is based on the Gaussian Mixture Model (GMM). However, the traditional GMM has the over-smoothed phenomenon and may get discontinuity of the converted speech due to the inaccuracy of the extracted feature information. In order to overcome it, we consider the dynamic spectral features between frames. The conversion function is also modified to deal with the discontinuities. The Speech Transformation and Representation using Adaptive Interpolation of weiGHTed spectrogram (STRAIGHT) algorithm is adopted for the analysis and synthesis process. Objective and perceptual experiments show that the quality of the speech converted by our proposed method is significantly improved compared with the traditional GMM method.

Comparing with orthogonal space time block code (OSTBC), quasi orthogonal space time block code (QOSTBC) can achieve high transmission rate with partial diversity. In this paper, we present a QOSTBC concatenated Reed-Solomon (RS) error correction code structure. At the receiver, pairwise detection and error correction are first implemented. The decoded data are regrouped. Parallel interference cancellation (PIC) and dual orthogonal space time block code (OSTBC) maximum likelihood decoding are deployed to the regrouped data. The pure concatenated scheme is shown to have higher diversity order and have better error performance at high signal-to-noise ratio (SNR) scenario than both QOSTBC and OSTBC schemes. The PIC and dual OSTBC decoding algorithm can further obtain more than 1.3 dB gains than the pure concatenated scheme at 10^-6 bit error probability.

In this paper, a new technique called color-SIFT model is devised for shot boundary detection. Unlike scale invariant feature transform model that uses only grayscale information and misses important visual information regarding color, here we have adopted different color planes to extract keypoints which are subsequently used to detect shot boundaries. The basic SIFT model has four stages namely scale-space peak selection, keypoint localization, orientation assignment and keypoint descriptor and all these four stages were employed to extract key descriptors in each color plane. The proposed model works on three different color planes and a fusion has been made to take a decision on number of keypoint matches for shot boundary identification and hence is different from the color global scale invariant feature transform that works on quantized images. In addition, the proposed algorithm possess invariance to linear transformation and robust to occlusion and noisy environment. Experiments have been conducted on the standard TRECVID video database to reveal the performance of the proposed model.

The paper provides a pre-stage of any data hiding method hiding message data in media data and a data extraction method of extracting the hidden data, wherein message data is dispersively hidden in digital media data, such as images, to prevent a third person from forging/modifying the message data easily. More specifically, the technique relates to a data hiding method in which media data is expressed as a media array while message data is expressed as a message array so that the array elements of the message array can be dispersively hidden in the media array randomly by scrambling order of particular array element of the media array based on a private key. It needs to declare that the proposed strategy aims only to enhance the watermark security. It is not used to improve the robustness of watermark.