The strategy of sequence image coding rate control by picture layer and mcroblock is proposed in this paper. Rate control on picture layer actually allocates a number of target bits for current frame to be encoded. Based on this number, a reference quantization parameter is selected on picture layer. In the course of determining the quantization step of every macroblock in a frame, the characteristics of an image itself should be fully considered. The characteristic of the image by the distribution characteristics of the DCT coefficients is denoted and a best quantization parameter for every macroblock in a frame is selected. The complexity of the algorithm is analyzed and an improved method is proposed. The experiment shows that the rate control strategy proposed in this paper can effectively reduce and avoid the overflow and underflow of the encoder buffer, and the output bit rates tend to stabilization. The peak signal-to-noise ratio (PSNR) of the reconstructed image is raised.

Monitor oriented RGB color spaces (e.g. sRGB) are widely applied for digital image representation for the simplicity in displaying images on monitor displays. However, the physical gamut limits its ability to encode colors accurately for color images that are not limited to the display RGB gamut. To extend the encoding gamut, non-physical RGB primaries may be used to define the color space, or the RGB tone ranges may be extended beyond the physical range. An out-of-gamut color has at least one of the R, G, and B channels that are smaller than 0 or higher than 100%. Instead of using wide-gamut RGB primaries for gamut expansion, we may extend the tone ranges to expand the encoding gamut. Negative tone values and tone values over 100% are allowed. Methods to efficiently and accurately encode out-of-gamut colors are discussed in this paper. Interpretation bits are added to interpret the range of color values or to encode color values with a higher bit-depth. The interpretation bits of R, G, and B primaries can be packed and stored in an alpha channel in some image formats (e.g. TIFF) or stored in a data tag (e.g. in JEPG format). If a color image does not have colors that are out of a regular RGB gamut, a regular program (e.g. Photoshop) is able to manipulate the data correctly.

The embedded multi-resolution coding in JPEG2000 make the ROI(Region Of Interesting) coding very successful. But it is processed in the wavelet domain, resulting in having to decompose some mature JPEG2000 baseline system coder and decoder to implement the ROI function. So a low computational complexity upscaling method in space domain, deduced from the relation of the original image and the image after wavelet transform, is proposed in this paper without modifying the baseline system. By the method, we should right shift the data outside the region of interesting, then use the normal JPEG2000 system to encode and decode the image, and subsequently we can get the image by left shift that data in the space domain after decoding. The method can get high performance quality in ROI in very low bit rate. And it has low computational complexity, needn't deduce the ROI mask in the wavelet domain. As Maxshift method in the wavelet domain, the Maxshift in space domain can be deduced and it decodes the compressed image without ROI mask. That is the ROI mask transmission is not need. The cost of the simpleness is that it will produce distinct tiring edge around the ROI. The method can implement the ROI function through simple extends with the JPEG2000 baseline hardware without the ROI function and it is significative in practice.

Three-dimensional image compression methods outperform their two-dimensional counterparts in the sense of higher rate-distortion performance for compressing volumetric image data. The state-of-the-art transform-based 3D compressors, such as 3D-SPIHT and 3D-DCT, are characterized for their rate control ability, where the qualities of the image, although are adjustable with respect to rates, are not explicitly controllable. A novel method, based on vector quantization in an enhanced image pyramid with error feedback, has been proposed, where the quality of the decompressed image only depends on the encoding of coefficients from the finest band and therefore a distortion-constraint transform coding is achieved. Compared to the previous image pyramid transform coders, its coding efficiency has been improved by using a cross-band classified vector quantizer (CBCVQ), where the encoding of current band will benefit from the encoding result from previous bands. Two explicit bit-allocation schemes, one is regarding the bit allocation across bands and the other is across the sub vector quantizers within each band, have been applied to minimize the total rate under the constraint of specified distortion. Evaluations have been performed on several data sets obtained by confocal laser scanning microscopy (CLSM) scans for vascular remodeling study. The results show that the proposed method has competitive compression performance for volumetric microscopic images, compared to other state-of-the-art methods. Moreover the distortion-constraint feature offers more flexible control than its rate-constraint counterpart in bio-medical image applications. Additionally, it effectively reduces the artefacts presented in other approaches at low bit rates and therefore achieved more subjective acceptance.

The registration of multi-sensor images is a basic and key technique in image fusion process. This paper presents two different methods aimed to targets with different characteristics. For the small target, the registration is completed by two steps. First, using imaging principle, with focus, resolution and pixel dimension of CCD, the Field of View (FOV) corresponding relation between different CCD is established, and then it is used to complete FOV registration by cutting and zooming out the FOV of target. Second, the method of using Maximum Close Distance (MCD) matching to register target position is proposed. For the extend target, the 6-parameters affine transformation model to describe the target in 3-D space is established, and the imperfection of popular used 4-parameters model is analyzed. Then using the model and a set of matched points, the optimal parameters are fit through least square method. At last, the condition of use 6-parameters model to border the 8-parameters perspective model is analyzed. The experimental results show these two methods are all efficiency, and can eliminate the difference of the different sensor images.

Digital image stabilization (DIS) is an elementary function of hand-held video-capturing devices. In conventional combination, video encoding is performed after DIS. However, both DIS and video encoding have the same time-consuming unit: motion estimation (ME). Thus, a novel methodology for fusion of DIS and video encoding is proposed in this paper to reduce the computation of ME and produce coding-compliant motion vectors (MVs). That is accomplished by re-using the MVs of DIS and the variable block sizes defined in H.264; thus, macro blocks (MBs) with MVs need not to compute MVs again. In addition to the MBs used in DIS, the rest area f a stabilized P frame is partitioned into many blocks by means of the variable block size as defined in H.264. By using the proposed method, the saving in chip area will result in great reduction of power and cost. According to experiments, it is approximated that 25% of the ME can be eliminated.

A novel multi-resolution image fusion method based on maximum directional contrast and area-based standard deviation is presented. First the source images are performed the multi-resolution wavelet decomposition and the directional contrast and area-based standard deviation are defined. And then, the wavelet coefficients for the fused image can be obtained by means of the fusion rules based on maximum directional contrast and area-based standard deviation. Finally, the fused image is reconstructed by inverse wavelet transform. By this scheme, the contrast and details from each original image are emphasized and enhanced in the fused image. The experimental results show that the fusion algorithm presented here is much effective.

This paper analyzes the low-level-light image and ultraviolet image's characteristics, and then the low-level-light and ultraviolet dual band false color preprocessing and fusion hardware system is put forward. To this system, real-time performance is an important factor. This system contains two parts. The first part is a FPGA+SDRAM architecture noise reducing system. The time domain average filter is applied to this part, because it meets the real-time requirement and can effectively decrease the low-light-level and ultraviolet image's flicker noise. The second part is the fusion system. Its core is the most advanced video processor TMS320C6711. The processor's EDMA can operate smartly to achieve the dual channel images' capturing, calibrating and false color displaying without the core processor's interference, while this trait is especially useful to the dual band image fusion system. In this part, for real-time performance consideration, the improved gray modulating fusion algorithm is used. The improvements aim at using the maximum potential of the core processor's architecture. This paper gives the hardware data flow of the time domain average filter algorithm, the image registration algorithm and the improved gray modulating fusion algorithm in detail, and the system's schematic is also included in this paper. This system achieves the low-level-light and ultraviolet image's noise reducing, and solves the worldwide problem, the image registration. And most important it is a real-time hardware processing system and can be easy to integrate and equip.

The goal of image fusion is to create new images that are more suitable for the purposes of human visual perception, object detection and target recognition. For Automatic Target Recognition (ATR), we can use multi-sensor data including visible and infrared images to increase the recognition rate. In this paper, we propose a new multiresolution data fusion scheme based on Daubechies Wavelet Basis (DWB) and pixel-level weights including thermal weights and visual weights. We use multiresolution decompositions to represent the input images at different scales, present a multiresolution/multimodal segmentation to partition the image domain at these scales. The crucial idea is to use this segmentation to guide the fusion process. Physical thermal weights and perceptive visual weights are used as segmentation multimodals. Daubechies Wavelet (at different levels) is choosen as the Wavelet Basis. Experimental results confirm that the proposed algorithm is the best image sharpening method and can best maintain the spectral information of the original infrared image. Also, the proposed technique performs better than the other ones in the literature, more robust and effective, from both subjective visual effects and objective statistical analysis results.

Flash movie is gaining widespread usage nowadays. There are dozen of Flash movies now available throughout the Internet. In order to allow Flash movies to be queried, we create an OXF (Object XML for Flash) model by converting Flash movies into XML format. We apply Structural Join Index Hierarchy (SJIH) indexing technique to enhance retrieval of OXF objects. A detailed cost model for query execution through SJIH is developed. Our experimental studies on using SJIH based on Flash movies has demonstrated performance advantageous when compared with other indexing techniques.

The paper brings forward a new method for identifying and annotating erotic films based on content analysis. First, the film is decomposed to video and audio stream. Then, the video stream is segmented into shots and key frames are extracted from each shot. We filter the shots that include potential erotic content by finding the nude human body in key frames. A Gaussian model in YCbCr color space for detecting skin region is presented. An external polygon that covered the skin regions is used for the approximation of the human body. Last, we give the degree of the nudity by calculating the ratio of skin area to whole body area with weighted parameters. The result of the experiment shows the effectiveness of our method.

This paper presents a novel approach for urban texture analysis. The approach applies the artificial immune theory in learning the texture filters for urban texture classifications. In this paper, urban textures are regarded as non-self, and non-urban textures are regarded as self. Texture filters are regarded as antibodies. The clonal selection algorithm is employed to evolve antibodies. Experimental results of urban texture analysis on aerial images are presented to illustrate the feasibility of the proposed method.

This paper presents a newly developed passive stereo display system. This system is based on normal PC computer, no special designed software and hardware is needed. The only special requirement is the graphics card should have dual video display connectors, that is, the video card has n-view functions. We use nvidia Quadro FX1100 graphics card in our system. We set the whole image resolution to 2048X768. By specially arrange the stereo images in this area, we set one of the screen to display the left image and the others displays the right image, through normal display driver software, we can get the separated stereo image pair displayed on the two screens. Output these two separated stereo pairs on two image projectors, and set a pair of perpendicularity polarizers in front of these projectors, a positive computer stereo image projection system is established. This system has the specification of simple, easy to construct, and hardware independent.

An AR (Augmented Reality) system can integrate computer-generated objects with the image sequences of real world scenes in either an off-line or a real-time way. Registration, or camera pose estimation, is one of the key techniques to determine its performance. The registration methods can be classified as model-based and move-matching. The former approach can accomplish relatively accurate registration results, but it requires the precise model of the scene, which is hard to be obtained. The latter approach carries out registration by computing the ego-motion of the camera. Because it does not require the prior-knowledge of the scene, its registration results sometimes turn out to be less accurate. When the model defined is as simple as a plane, a mixed method is introduced to take advantages of the virtues of the two methods mentioned above. Although unexpected objects often occlude this plane in an AR system, one can still try to detect corresponding points with a contract-expand method, while this will import erroneous correspondences. Computing homography with RANSAC algorithm is used to overcome such shortcomings. Using the robustly estimated homography resulted from RANSAC, the camera projective matrix can be recovered and thus registration is accomplished even when the markers are lost in the scene.

Accurate measurement of blood vessel diameter on ultrasonic images is a potential part in diagnosis and physiological study. In this paper, a method of blood vessel diameter measurement has been presented. In our approach, we firstly used a histogram equalization technique to enhance the contrast of ultrasonic images. Secondly, a Guassian filter is used to remove the noisy pixels while preserving the important information. Next, we applied Haar wavelet filter to extract the blood vessel shape. Moreover, we used the nonmaxima suppression and thresholding techniques to get the blood vessel shape in the binary format. Finally, the users have to drag a mouse to measure the diameter of blood vessels. From the results, we found that our approach yields promising results compared with conventional methods and other mother wavelets. An ultrasonographer also confirmed the results. Therefore, our approach leads to an effective method for blood vessel diameter measurement.

A novel method providing a supervised processing of medical image for segmentation is presented. This method was based on a pyramid-structured wavelet-transform and improved watershed transform algorithm. The method contains three consecutive stages: image segmentation based on multi-resolution watershed transform, region projection and mergence with extracted multi-future information, edge refinement based on fuzzy information fusion. In the processing, both texture and gray variation information are used inside the tissue regions, and only gradation information is used near the edges of regions. Experimental results for the proposed algorithm indicate feasibility and reliability for certain medical images segmentation.

CT scans are thin cross-sectional, radiographic images that can be obtained at any body level. CT images can describe the soft tissues with better clarity because it is more sensitive to slight differences in attenuation than standard radiography. Image segmentation is the key process to identify body fat in CT images. CT images at different body levels have different structures and hence different grayness histogram. Furthermore, the grayness histogram itself, in one CT image, has multiple peaks. Therefore, three segmentation methods, automatic threshold segmentation, morphological reconstruction segmentation, and potential function clustering segmentation, are used in this paper. Body fat contents and distributions are got according to segmented CT images. Experiment results show the effectiveness and stability of the multi-thresholds image segmentation method based on potential function clustering.

Digital optical microscopy, coupled with parallel processing and a large arsenal of labeling techniques, offers tremendous values to localize, identify, and characterize cells and molecules. This generates many image informatics challenges in requiring new algorithms and tools to extract, classify, correlate, and model image features and content from massive amounts of cellular and molecular images acquired. Image informatics aims to fill this gap. Coupling automated microscopy and image analysis with biostatistical and data mining techniques to provide a system biologic approach in studying the cells, the basic unit of life, potentially leads to many exciting applications in life and health sciences. In this presentation, we describe certain new system biology applications enabled by image informatics technology.

Considering the fact that the remote sensing image is mainly captured by a linear CCD with the push-broom way which the image varies over time, the time-varying wavelet packets for image restoration is proposed in the paper. On one hand, the result of the method is that the problem of the correlation between the images is solved, and on the other hand, it is the method that can remarkably reduce the calculating overhead and the data throughput, which is a key innovation for a real-time system. In this paper, the optimized wavelet packet bases by double tree searching algorithm are adaptively changed in different time. To realize the algorithm, we presented a dual DSPs real-time parallel system. The parallel system based on TMS320C6416-7E3 DSP has the characteristics of modular and flexible design and maintainability. The ping-pong structure and the streamline structure are both designed in the system. According to the complexity of the algorithm and the requirement of the data throughput, one of the two parallel structures can be realized freely only by changing a bit. It can realize the restoration algorithm with 4096*4096 images in real-time by demonstrated by our experiment in practice.

For video communication over Internet, restricted by transmission bandwidth, the video sampling rate is much lower than the Nyquist frequency decided by the features of sampled target. So, undersampling is a common case. Undersampling produces aliasing noise within sampling passband, which generates small-scale artifacts all over the image. Aliasing noise is much more severe factor than other noises in sampled imaging system which degrade the quality of image. Usually, chrominance components suffer from aliasing noises much more heavily than luminance component, because their sampling rate is a half of the one of luminance component. We propose novel sampling data processing methods and develop the algorithm derived by Zhi Kuan Chen[1]. In our sampling mechanism, we don't directly reduce the pixel density of chrominance components, instead, we construct 4 low density frames from one input high density frame. These four frames have different spatial shifts relative to original frame. We apply the algorithm to these four frames. The aliasing noises are reduced or eliminated. Effect of our aliasing noises processing methods depends on the accuracy of spatial shifts measuring. Fortunately, in our mechanism, the displacements or phase shifts can be decided precisely

The Mumford-Shah model has been well acknowledged as an important method for image segmentation. This paper discussed the problem of simultaneous image segmentation and smoothing by approaching the Mumford-Shah paradigm from a numerical approximation perspective. In particular, a novel iterative relaxation algorithm for the numerical solving of the Mumford-Shah model was proposed. First, the paper presented mathematically the existence of a solution in the weak formulation of GSBV space. Second, some approximations and numerical methods for computing the weak solution were discussed. Finally, a minimization method based on a quasi-Newton algorithm was put forward. The proposed algorithm found accurately the absolute minimum of the functional at each iteration. Considering the important role of a discrete finite element approximation method in the sense of Γ-convergence, an adjustment scheme for adaptive triangulation was applied to improve the efficiency of iteration. Experimental results on noisy synthetic and jacquard images demonstrate the efficacy of the proposed algorithm.

The most obvious difference between digital photofinishing and other color reproduction devices, such as printer, is its exposure mechanism. So the color control system is based on the analysis of exposure characteristic. To offset the colorimetric shift by developer, three target density curves are proposed and applied to the pre-adjustment of exposure mechanism. Since the color control system is considered static via pre-adjustment, it can be modeled by the color space transform from colorimetric space to dye space. By comparison, the neural network method is utilized to realize this transform. To get uniform samples in CIE Lab color space for network, a cumulative dye slicing method is designed. At last, the GretagMacbeth ColorChecker color rendition chart is printed, and the results indicate that this color control system achieves the color difference requirement of high-quality reproduction.

The problem of stabilization of the image sequence from a camera mounted on a moving platform such as airborne reconnaissance is considerably more complex. Motion composed of not only translation but also rotation is a very common. For the long focus, long-range imaging system, to estimate and compensate rotation movement between the frames is very important. Through analyzing the effect of rotation on the quality of video sequence, the paper proposes an effective algorithm applied to the rotation and translation movement estimation. This algorithm is composed of three parts: Firstly, the image is divided to many regions and make sure that every pixel motion in each region is uniform; Secondly, the translations vector of each region are obtained by representative point matching technique. Finally the globe motion vectors (include translation and rotation) between the reference frame and the current frames is obtained by solution of a set of linear equations which come form a given affine motion model. Compensation method of motion among the frames is also an important factor to affect the stability of image sequence. If the image sequence is compensated directly with the acquired movement vectors, it will feel the jumping among frames. In the paper, the parameters to compensate for the current image have be determined by average value filtering and been applied to given motion model to stabilize the image stream. It can realize smoothly processing to assure the image sequences stabilization.

The performance of direct viewing low light level (LLL) imaging system is mainly determined by three factors: photons noise, MTF of optical system(OS) and human eyes characteristic. And the image detecting theory which denotes the optimal performance of imaging system has been a positive impetus for the development of the LLL imaging and night vision technique. The system minimum resolvable angle was traditionally used to estimate the image detecting performance which is mainly determined by photons noise at low target illuminance and by MTF at high target illuminance. This criterion can represent the system performance on the whole; however, assuming the signal to noise ratio (SNR)of the image and MTF of OS uncorrelative, is theoretically not complete, since the two factors interrelate actually. From the viewpoint of signal response, the MRC (minimum resolvable contrast) model of the ideal direct viewing LLL imaging system was deduced on the basis of human eyes characteristic. It is a more comprehensive evaluation method for imaging system performance, and can combine with the forecasting model of operating distance to analyze the general performance of night vision system. In conclusion, the relationship and the difference between the MRC model and the traditional detecting equation were investigated.

This paper presents a new automatic image segmentation method for segmenting moving object in complex environment by combining the motion information with edge information. We propose an adaptive optical flow method based on the Horn-Schunck algorithm to estimate the optical flow field. Our method puts different smoothness constraints on different directions and optical flow constraint is used according to the gradient magnitude. Canny edge detector can obtain the most edge information but miss some pixels. In order to restore these missing pixels the edge has a growing based on the continuity of optical flow field. Next, by remaining the block that has the longest edge could delete the noise in the background, and then the last segmentation result is obtained. The experimental result demonstrates that this method can segment the moving object in complex environment precisely.

Several new super-resolution image restoration algorithms based on orthogonal discrete wavelet transform are proposed, by using orthogonal discrete wavelet transform and generalized cross validation ,and combining with Luck-Richardson super-resolution image restoration algorithm (LR) and Luck-Richardson algorithm based on Poisson-Markov model (MPML). Orthogonal discrete wavelet transform analyzed in both space and frequency domain has the capability of indicating local features of a signal, and concentrating the signal power to a few coefficients in wavelet transform domain. After an original image is "Symlets" orthogonal discrete wavelet transformed, an asymptotically optimal threshold is determined by minimizing generalized cross validation, and high frequency subbands in each decomposition level are denoised with soft threshold processes to converge respectively to those with maximum signal-noise-ratio, when the method is incorporated with existed super-resolution image algorithms, details of original image, especially of those with low signal-noise-ratio, could be well recovered. Single operation wavelet LR algorithm(SWLR),single operation wavelet MPML algorithm(SW-MPML) and MPML algorithm based on single operation and wavelet transform (MPML- SW) are some operative algorithms proposed based on the method. According to the processing results to simulating and practical images , because of the only one operation, under the guarantee of rapid and effective restoration processing, in comparison with LR and MPML, all the proposed algorithms could retain image details better, and be more suitable to low signal-noise-ratio images, They could also reduce operation time for up to hundreds times of iteratives, as well as, avoid the iterative operation of self-adaptive parameters in MPML, improve operating speed and precision. They are practical and instantaneous to some extent in the field of low signal-noise-ratio image restoration.

This paper proposes a new tree-like fuzzy binary support vector machines multi-class classifier (FBSVM) for the optical character recognition task. We construct this tree-like classifier by fusing of fuzzy clustering technique and support vector machine (SVM). In k-class task, the new classifier contains k-1 SVM sub-classifiers, but the "one-against-one" method which is usually used contains k(k-1)/2 sub-classifiers. This method also overcomes the drawback such as unclassifiable region that the "one-against-one" method has, and has a good classification performance. Furthermore, it needs less memory. By applying the new classifier to the real mail zipcode digits recognition task, the experimental results indicate that the FBSVM has a better recognition performance.

Real-time target detection against strong (bright) background under daytime is a challenging and leading edge subject, and also is a key technique for imaging tracking system. Strong background makes CCD image sensor work in critical saturation state, and imaging target contrast is very low. It's very difficult to accurately and stably track due to the complex characteristics of imaging target, such as strong clutter background, low contrast, and low signal to noise radio (SNR). So the key techniques for detecting and tracking target are eliminating the disturbance of diffuse reflection and beacon, synchronous detection, improving the performance of real-time image processing with high frame rate and high sampling rate. A robust strategy for detecting and tracking day-time target was proposed in this paper. A series of efficient approaches ware presented to improve performance of detection and tracking in precision and stability, including strong background and noise suppression, image enhancement, adaptive thresholding, region merging based on morphology, recognition and tracking algorithm and so on. In the end, we summarized and built the effictive flow for detecting and tracking target against strong background under daytime. The results of combining computer simulation with practical detection experiments show that the above-mentioned approaches are feasible and significant for real-time tracking system.

A Number Plate Recognition System with a special linear CCD imager is presented in this paper. The system is fixed on the top of the roadway to obtain the images of the moving vehicles. With the linear image sequence, we have detected the position of Number Plate with corresponding algorithm and output 2-dimension sub-image of Number Plate by standard video for Optical Character Recognition (OCR). The 2-dimension image combined with the linear image sequence has higher transverse resolution (2048pixels) than that of the normal standard video (768pixels). Moreover, even images and accurate location of the characters can be gained with the linear CCD imager. The system is based on FPGA technology, and it can detect the vehicles Real-timely and recognize Number plate.

On-line real-time detection method for the defect of TFT-LCD is becoming increasingly important as TFT-LCD has replaced CRT displays and become the first choice in many applications. Traditional defect inspection methods of TFT-LCD are based on clear features and exact mathematic models. However, the defects of TFT-LCD are of strong complexity and vagueness. Moreover, determining the defects is a complicated process, which is influenced by the objective characteristics of the defects as well as the subjective factors of the observer. Therefore, it is very difficult to establish the accurate mathematical models for the defects. A fuzzy expert system approach is proposed for the defect inspection of TFT-LCD. Tests indicate that this system could emulate the experts or experienced operators to realize the automatization of the defect inspectin of TFT-LCD.

A face detection method based on higher order statistics is proposed in this paper. Firstly, the object model and noise model are established to extract moving object from the background according to the fact that higher order statistics is nonsense to Gaussian noise. Secondly, the improved Sobel operator is used to extract the edge image of moving object. And a projection function is used to detect the face in the edge image. Lastly, PCA(Principle Component Analysis) method is used to do face recognition. The performance of the system is evaluated on the real video sequences. It is shown that the proposed method is simple and robust to the detection of human faces in the video sequences.

It is the key problem to miss distance measurements that how to obtain Doppler frequency from radar echo signals, the measurement accuracy of Doppler frequency affects immediately estimation accuracy of miss distance parameters acquired by non-linear optimize arithmetic. A new method is developed to extract Doppler frequency from the image in this paper. In this method, the radar echo signals, which are received by antenna arrays, are transformed to time-frequency domain through Short Time Fourier Transform (STFT) firstly, then the gray-scale image can be obtained that make the intensity of the spectrum as the gray value, further the centroid estimation method is applied to obtain the extraction of Doppler frequency. Simulation results show that the proposed method can effectively extract the Doppler frequency in the low signal noise ratio and get the exact parameter estimation of scalar miss distance.

Smoke control is one of the important aspects in atrium fire. For an efficient smoke control strategy, it is very important to identify the smoke and fire source in a very short period of time. However, traditional methods such as point type detectors are not effective for smoke and fire detection in large space such as atrium. Therefore, video smoke and fire detection systems are proposed. For the development of the system, automatic extraction and tracking of flame are two important problems needed to be solved. Based on entropy theory, region growing and Otsu method, a new automatic integrated algorithm, which is used to track flame from video images, is proposed in this paper. It can successfully identify flames from different environment, different background and in different form. The experimental results show that this integrated algorithm has stronger robustness and wider adaptability. In addition, because of the low computational demand of this algorithm, it is also possible to be used as part of a robust, real-time smoke and fire detection system.

Motion estimation (ME) and compensation (MC) is critical to the performance of an encoder, because the procedure is computationally intensive. To reduce the calculation, people work out some kinds of fast search algorithms for motion estimation, and dramatically improve the performance. This paper uses the Intel Pentium CPU's MMX, XMM registers and some Single Instruction Multiple Data (SIMD) instructions to accelerate the calculation, especially, uses PNI (Prescott New Instruction). We could load more pixels' values to a register at the same time. With PNI’s instruction LDDQU, we could load 16 bytes to XMM register even they cross a cache line boundary. Therefore, we could calculate (add, subtract, average, get absolute differences) multiple samples in a single operation. The parallel operations will significantly increase the speed of the ME and MC, irrespective of which kind of search algorithm.

The images taken by digital camera can not only be filmed out or printed out colorfully but also shown on televisions and computers. In order to output a higher quality image, analyzing and researching of digital images by software and algorithm were given in this paper. For example, researching the resolution of digital images can provide the size for optimize the output picture, make the images size after computer processing had the same detail as the original one with higher sizes, the images after processing could gain higher magnification and clearer interface; researching the color of the images after computer processing could correct the color which was out of line, the output images could be more verisimilitude; researching the whole image through optimal processing at deficient parts could decrease the noises which caused by the digital camera system; analyzing objects of the images, such as edges, color and luminance changing, with interrelated algorithm of digital image processing could make the images super-resolution zooming more perfect, image cutting, image mosaic, print character of the images were also discussed.

This paper proposes a method using exploring agent for the noise elimination and crack recognition in binary images which originate from the objective gray level images. A mean filtering method is introduced to correct non-uniform background illumination and obtain the dynamic thresholds, which are used to convert the original 255 scales gray level image into binary images. The pavement crack figures in the binary image have been contaminated by randomly distributed noisy dots, and in most cases, the crack shape and orientation can't be represented by specific functions. The exploring agent method using sense-compute-act loop, presented in this paper, can be employed to determine the crack and eliminate the random noise. The exploring agent and the Least Square Fit (LSF) method separately have unique characteristics in recognizing the crack intersection and orientation, and automatically running along the crack. The traces of the exploring agent are the skeleton of the pavement crack, and the number of steps can be used to calculate the length of the crack. The sense, compute, and act ability of the exploring agent iterate to guarantee the effect in processing randomly distributed features of image during the actual processing.

Interest in color appearance models (CAM) has been greatly stimulated recently by the need in handling digital images. This article demonstrates that a multi-layers feed-forward artificial neural network with the error back-propagation algorithm was used to approximate color appearance model CIECAM02 with different white points and different media. For the prediction of the forward and inverse model respectively, in order to realize accurate mapping, especially to the inverse model, color spaces conversion between input color space and output color space (that is cylindrical coordinates and rectangular coordinates) was implemented before training the neural networks. Meanwhile we approximated the combination of the forward and inverse CIECAM02 models employing a neural network for different conditions including whites (D65 or D50) and media (booth and CRT) in order to realize the color transformation from one medium to another conveniently. The experimental results indicated that the prediction could satisfy the accuracy requirement. So in practice we can choose these two kinds of different prediction ways to meet our need according to different situations.

It is often necessary to follow the axial movement of a micron particle, such as the one trapped in optical trap, in addition to its radial movement. A new method based on the information entropy is developed for measuring the axial displacement, which is then used to reconsider the drag force method for measuring the radial trapping force and stiffness in an optical tweezers system. It is found that the new equilibrium position of the bead displaces not only radially but also axially when the surrounding viscous fluid flows at constant lateral velocity. The result implies that the trap stiffness measured in such a way is not really for the same horizontal plane. In addition, the measured trajectory of the bead (both radial and axial displacements) shows that the sphere escapes from the optical trap upward in stead of radially when the fluid velocity reaches the critical value. The fact indicates that the escape force is not the maximal radial trapping force as commonly accepted. It also deduced that the axial movement of the bead is one of error sources for trapping force calibration using the drag-force method.

In order to hide secrete information in remote sensing image, we proposed an algorithm for secrete information hiding which was adaptive to the feature of remote sensing image. Firstly, we segmented and extracted the secrete information in remote sensing image, and made supplement of gray values in the area corresponding with the secrete information and then produced the disguised remote sensing image which was wiped off secrete information. Then we used for reference the idea of digital watermarks and feature of HVS (Human Visual System) and embedded the secrete sub-image imperceptibly and adaptively into the disguised remote sensing image to produce the disguised remote sensing image in which there hid secrete sub-image. In addition, during the course of extracting secrete information and resuming the remote sensing image, this algorithm didn’t need the original remote sensing image and was a blind one. To those algorithms for information hiding, imperceptivity and amount of hidden information are the most important and robustness is less. And experimental results show that this algorithm is not only quite transparent and has a good effect for large amount of secrete information hiding, but also has a strong robustness against such image attacks as JPEG lossy compression, median filtering, noise adding, scaling, cropping and rotation. Furthermore this algorithm has no influence on such applications as edge detection and image classification of the disguised remote sensing image which has been hidden the secrete information.

Dielectric barrier discharge is a novel system for studying the pattern formation. In this system, spiral pattern has been observed for the first time. For the element of the pattern emits itself, the image can be obtained to give some information about the behavior of the spiral pattern. The image of spiral pattern is analyzed based on the software Matlab6.1. The spatial distribution of the intensity in the image gives a normal spiral structure. The temporal distribution of the intensity shows that the spiral pattern has a periodical behavior in a second time scale. All of the results are consistent with the theoretical simulations.

Architecture of an IP-based narrow-band videophone system is proposed in this paper for convenient videophone calls between any two computers even if being placed in two different LANs within network agents. The bandwidth need of each call is less than 256 kbps. The system consists of two kinds of entities: Videophone Terminals (VPT) and a Video Call Server (VCS). A VPT is actually a microcomputer program, composed of 4 primary parts, an audio codec, a video codec, a media deliverer/receiver and a call controller. The basic functions of the VCS include videophone number generation and management, access admission and address resolution. The VCS with a public IP address plays an important role in the system especially when a video call has to penetrate through network agents. Each VPT in the system gets its own external transport address from the VCS through registration process. A calling VPT would receive the external transport address of the called VPT from the VCS through address resolution. The proposed system works and is helpful to accelerate the realization of people's videophone dream over IP networks.

CRT color gamut boundaries can be determined by two steps workflow. Firstly, the display should be calibrated with the method recommended by CIE to characterize the relationship between CIE tristimulus values and DAC values. The nonlinear relationship of each electronic channel between the color of the radiant output of CRT displays and the digital DAC values can be characterized accurately with GOG model using parameters of gain, offset, and gamma. Secondly, color gamut boundary can be determined using a fast and accurate algorithm. Generally, in a color space, any chosen degree of lightness will reduce that space to a plane. The color gamut on this equal-lightness plane can be transformed into RGB DAC value space. Since locations on the edges and surfaces of RGB DAC value space will correspond colors with relatively high saturation, the boundary of the curved surface in RGB DAC value space can be quickly computed for certain lightness. The accurate color gamut is obtained by mapping this boundary over to such a perceptual color space as CIELAB or CIELUV uniform color space. The key issue of this algorithm is to compute the equal-lightness curved surface in RGB DAC value space. The resolution of device gamut description depends on the number of segments that the lightness axis is separated into in the perceptual color space.

The Paper is the study of Gabor wavelet neural network algorithm and its application in gray image target recognition. The mostly thought t are real time recognizing gray image target with Gabor wavelet neural networks algorithm. The main thoughts are through combing the forward neural networks (BP net) with Gabor wavelet based on they were applied in target feature extraction and recognition. A model of Gabor wavelet neural network is constructed with automatic target recognition, the good impact is gained when it is applied target recognition. The principle of Gabor filter is expounded. The multi-channel Gabor filter is designed based on theory and practicality, the neural network recognizing algorithm based on multi-channel Gabor filter feature is presented. Training algorithm of Gabor wavelet neural networks model was given out. Principally analyzed Gabor wavelet neural networks from theory, in the mean time training algorithm of Gabor wavelet network suited to target recognition was designed by BP algorithm. Theory and simulate experiment indicated the astringency and robustness of this algorithm excelled BP net. Target was recognized by this algorithm not only increased recognition precision but also overcame the bug of BP algorithm get in minimum

High-speed relative motion between the imaging system and the object scene in the duration of the exposure will lead to a radiate blur from image center to image edge, which reduces effective angular field. The captured images are blurred more seriously with the increase of relative motion velocity. The blurred image makes object detection and recognition extremely difficult. The purpose of this paper is to give a degradation model for high-speed relative motion blur, and to analyze the degradation of special circular symmetric images, and factors that affect radiate degeneration, including relative motion velocity, distance between target and camera, and frame frequency. The proposed model explains the phenomenon that the farther from the image center, the more serious the blur becomes. Based on theoretical analysis, the degradation model of practical discrete imaging process is studied. Blurred gray values of the discrete pixels can be obtained along the blurring paths, which are a series of lines pointing at the image center. The radiate degenerate image is simulated finally.

The scalable image coding is an important objective of the future image coding technologies. In this paper, we present a kind of scalable image coding scheme based on wavelet transform. This method uses the famous EZW (Embedded Zero tree Wavelet) algorithm; we give a high-quality encoding to the ROI (region of interest) of the original image and a rough encoding to the rest. This method is applied well in limited memory space condition, and we encode the region of background according to the memory capacity. In this way, we can store the encoded image in limited memory space easily without losing its main information. Simulation results show it is effective.

In recent years, along with the development of computer technology, compression technology and network technology, how to efficiently transmit multimedia information over network has become a main concern by industry and research institution. MPEG-4 FGS technology is introduced and the rate allocation optimizing problem during MPEG-4 FGS bitsream streaming over network is analyzed. In order to performing rate allocation, the rate-distortion function of FGS enhancement layer should be known, so this paper firstly analyzes the rate-distortion function. Secondly, several objectives that the rate allocation should meet are pointed out. Finally, two rate allocation optimizing models are proposed: constant distortion model and minimum distortion model, and the corresponding mathematic expressions are given. The result revealed that the two optimizing models are different, and they can be selected according to different network variance conditions.

Region-of-interest (ROI) image coding is an interesting feature in JPEG2000, which allows for encoding the ROIs in an image with better quality than the background (BG). Considering the limitations of two standard ROI coding methods, a new ROI coding method called generalized partial bitplanes shift (GPBShift) is presented in this paper. To control the relative importance between ROI and BG, the method divides the bitplanes of ROI and BG coefficients into two parts by using scaling values S1 and S2, respectively. Instead of shifting the bitplanes all at once by the same scaling value S in the standard methods, GPBShift shifts part of them on the basis of the bitplane shifting scheme. The GPBShift not only can code arbitrarily shaped ROI without explicitly transmitting any shape information to the decoder, but also flexibly select the scaling values to adjust relative compression quality in ROI and BG. Additionally, the method can efficiently code multiple ROI with different priorities in an image. Experimental results show that the GPBShift method can provide significantly better visual quality than the Maxshift method at low bit rates, and higher coding efficiency than the general scaling based method.

In this paper, an IP-based video lab-monitor system is proposed in order to efficiently supervise and manage the Electrical Engineering Example Lab Center of Hubei Province. The proposed system is composed of one Control & Display Unit (CDU) and a number of Lab View Units (LVU). The CDU is placed in the lab-supervisor’s office, while each LVU with a video camera is placed in one of the labs to be watched. The CDU and all LVUs are connected with an IP network. An LVU is mainly composed of 4 parts: Video Capture, Video Encoder based on H.263, Media Deliverer and Communication Controller. Accordingly, the CDU is composed of the following parts: a Center Controller, a Media Receiver, a Multi-Video Decoder and a Multi-Video Displayer. The supervisor can simultaneously watch the dynamic scene of 16 (4x4) labs on the CDU, with a resolution of 176 x 144 for each lab. He may choose to watch 4 (2x2) labs or only one lab at a time with higher resolution.

Motion estimation plays an extremely important role in video coding. The objective of the motion estimation is to remove the temporal redundancy between video frames so that the video sequences can be coded efficiently. In this paper, an improved fast motion estimation algorithm, based on the successive elimination algorithm (SEA) of Li and Salari, is studied. This fast motion estimation algorithm results in the same displacement vectors as the exhaustive search algorithm (ESA) with a reduced computational load. An improved fast motion estimation algorithm, introducing further computational load reduction with negligible distortion, is proposed, and a transform coder based on the improved algorithm is developed. Implementation issues are discussed and compared. Experimental results show that the number of searching operations can be reduced dramatically with the help of the fast motion estimation algorithm.

The raw data in binocular stereo image sequences is twice as that of monocular images, the large amount of information should be reduced. As a result there has been increasing attention given to image compression methods specialized to stereo pairs. Much of this work has concentrated on improving the disparity compensation process and codes the residual image similarly to a monocular image where one view is used to predict another, and the difference is coded. The residual image is usually composed primarily of strong vertical direction edge components surrounded by large areas of near zero intensity. The residual images have different characteristics, but they behave uniquely statistical regularity. This property is demonstrated experimentally in the paper. Two interested statistical variables are described, the one is the total number (N) of the pixels with near zero intensity in the residual image and other is the coordinate displacements (Δx, Δy) between the left and right image frames for get the residual image. Experimental results indicate that the curve between the parameters N and variables (Δx, Δy) may be fit by Gaussian function. The maximum of the variable Nm corresponding to the optimal displacements (Δxop, Δyop) may be estimated by the Gaussian approximation. An algorithm is further provided to quickly predict the minimal redundancy of the residual image and the corresponding displacement. It is shown how such characteristics may be of great benefit to quickly achieve the higher compression ratio.

Taking into account the lack of prior-information on account of single image, we design a image super-resolution reconstruction method, based on the possibility and the theory basis of the single image super-resolution reconstruction. Analysis of the process of the algorithm is also included. Because the theory and method of Markov random field are being developed constantly, the theory and method may described the part statistic of the image. The paper analyses the image super resolution reconstruction based on Markov Random Field, this apply the image super resolution processing. This presented a super resolution reconstruction technique based on wavelet decomposition and Markov Random Field, and has carried on the experimental research. Experiment result proves that the super resolution processing method on the basis of the Markov random field (MRF) can obtain the good super resolution restoration processing result.

A new method to deal with images by computer is put forward, which is more convenient for the eyes to identify and much easier to understand. As the image edge is a basic character of image, checking it is one of the most important parts in processing the image. The traditional technique is to use the edge detection algorithm, which is to detect the gray level changes of every pixel of image in some epsilon, and to detect the image edge by using the changing regular of directional derivative. But sometimes there is uncertainty of image edge, and man can't distinguish it is the edge or not. Diagram algorithm can't solute this pivotal problem. In order to turn the fuzzy edge to be in focus and solve the problem above, this paper mentions fuzzy enhancing technique to realize image edge's being detected. Fuzzy technology is a newly rising technology used in many fields, especially in the image domain, and fuzzy enhancing technique is one important portion of the fuzzy technology. Based on this technology, this paper firstly sets the image fuzzy characteristic plane of original image, secondly proceeds the fuzzy enhancement, and then detects the edge by Sobel differential arithmetic. At the end of the paper, it realizes the histogram algorithms and the fuzzy enhancing algorithm by Visual C++. Results of the experiment show that fuzzy enhancing algorithm is a superior one in image procession.

In the research of super resolution reconstruction based on a set of images, we put emphases on analyzing the algorithm of super resolution based on multi-frame image or image set, i.e. how to reconstruct super resolution image using several low resolution images and under-sampled images. Discussion on recursively iteration reconstruct super resolution image is carried out, under both cases with noise and without noise. In this paper we bring Kalman filter theory into image super resolution rebuilding algorithm, make use of Kalman filter to go on a movement estimation to sequence image, and present a kind of simplified image super resolution reconstruction algorithm based on recursion iterative movement estimation of Kalman filte, and analyze this method. Then we make use of advanced method to test standard sequence image and acquired sequence image by shooting, and get a good result.

The relationship between gray value of pixels and macro-information in image has been analyzed with the method in statistical mechanics. After simulating and curve fitting with the experiment data by statistic and regression method, an adaptive threshold model between average gray value and image threshold has been proposed in terms of Boltzmann statistics. On the other hand, the image characteristics around the eye region and the states of eyeball also have been analyzed, and an algorithm to extract the eye feature and locate its position on the image has been proposed, furthermore, another algorithm has been proposed to find the iris characteristic line and then to coordinate the iris center. At last, considering the cases of head gesture, different head position, and the opening state of eyes, some experiments have been respectively done with the function based on the adaptive threshold model and the designed algorithms in eye-gaze input human-computer interaction (HCI) system. The experiment results show that the algorithms can widely be applied in different cases, and real-time iris tracking can be performed with the adaptive threshold model and algorithms.

A wavelet-based approach of aerial blurred image restoration is proposed in this article. Image motion is inevitable in photographing for aerospace camera. Though some Image Motion Compensation(IMC) schemes are applied in aerial imaging system, the ultimate image will be blurred in certain extent for the existence of IMC residual error, while the forward image motion is the key element among all image motions which lead to image blurring. First the course of blurring caused by forward image motion is expressed using wavelet transform, and a multiresolution sparse matrix representation of the degeneration model is obtained according to the wavelet transform. Subsequently a regularizing restoration algorithm is deduced from it, and which can smoothly restraint the processed result efficiently. In the end the proposed approach is tested in MATLAB. The blurred image is restored using above-mentioned wavelet algorithm, conventional contrary filter and Wiener filter algorithm separately. The conclusion that the wavelet-based restoration algorithm is superior to other two approaches is obtained by comparing the restored image’s value of mean gradient. The calculating quantity of the wavelet-based blurred image restoration approach isn't large and it has good practicability in the field of image interpretation and aerial survey or drawing.

Computer Generation Image (CGI) system is main part of the flight simulator. Usually, the Image Generation Workstation to be used in CGI system is very expensive. This paper presents a low cost CGI system based on PC and we use it to develop a flight simulator, which has a 3 channels display system with wide field-of-view (horizontal 240°× vertical 60°). First, the paper discusses the architectural principles of CGI system and provides a CGI system based on PC. Second, a type of 3 channels display system is described. The display system is designed in folded flat screen using rear project curtain. Adjacent screen shapes a 100° corner and each channel provides 80°×60° field-of-view. Furthermore, some key technologies, such as image connection, channels synchronization and channels communication, are discussed in detail. At the end of paper, a flight simulator with 3 channels visual scene is described and it is successfully implemented in training pilot. PC-Based CGI system also can be applied to a lot of fields including multimedia commerce, medical media databases visualization, CAD/CAM, scientific computing visualization, and entertainment.

Advanced observing and collimating technology demands excellent image detectors. While improving the performance of single image detector, methods of extending the observing ability through new pattern based on existing image detectors should be studied. Multi-band image color fusion bends itself to synthesize natural scene color image through different band mono-chromic information [1]. Real-time image matching is the choke point of this technology. The accuracy and validity of traditional image correlation-matching algorithm are degraded by the presence of the gray change, object variation and image noise. For it uses the method of adding the difference of the pixels' gray value according to the corresponding position as degree of mismatching. In addition, natural scene different band mono-chromic image spectrum respond characteristics vary sharply. We studied image features of low-light and infrared images; founded the arithmetic model of different band image matching; put forward image margin correlation matching based on margin pixel detecting; and realized it on Altera EP1S80 developing system. The result shows that this new algorithm is effective to different band image matching.

Two-dimensional (2-D) digital filters are widely useful in image processing and other 2-D digital signal processing fields,but designing 2-D filters is much more difficult than designing one-dimensional (1-D) ones.In this paper, a new design approach for designing linear-phase 2-D digital filters is described,which is based on a new neural networks algorithm (NNA).By using the symmetry of the given 2-D magnitude specification,a compact express for the magnitude response of a linear-phase 2-D finite impulse response (FIR) filter is derived.Consequently,the optimal problem of designing linear-phase 2-D FIR digital filters is turned to approximate the desired 2-D magnitude response by using the compact express.To solve the problem,a new NNA is presented based on minimizing the mean-squared error,and the convergence theorem is presented and proved to ensure the designed 2-D filter stable.Three design examples are also given to illustrate the effectiveness of the NNA-based design approach.

Projection displays are widely applied as tools for multimedia in conference room presentation, education center, R&D center and more places. To provide a more interactive environment, a new kind of human-computer interactive device is designed and presented. A two-dimensional CCD is the sensor of the unit. Through optical filter, CCD exports full video signal including a series of isolated positive pulse caused by the specific light-spot target generated from a specific light-pen. Through a video sync separator, combinational logic and sequential logic process of the full video signal, the target image's two-dimensional position on the light sensitive layer of CCD can be gained. The specific light-pen also sends the function logic message to the controller part through wireless communication. A microcontroller will combine the position information and function message, and then send it to computer through RS-232 of USB interface. The software in computer will process these messages. The specific light-spot's relative coordinates in the projection screen is gained. With the coordinate and the function message, the software will drive the computer to implement certain functions. With the specific light-pen, one can control the computer, take notes and shape his desire in the screen. Now the device is applied in LCD projection displays and it also can be applied in any large screen display. With the improvement of the system and the software, the function will be more powerful and provide a more interactive human computer interface (HCI).

This paper presents a novel method for speckle reduction in ultrasonic images. Firstly, a particular filtering kernel is defined by decomposing the local rectangular neighborhood into asymmetric sticks pointing outside with variable orientation from the investigated pixel. Then the local mean and variance along each stick are calculated using a template based convolution algorithm. Finally, a pseudo-diffusion model is derived to diffuse the intensity averages of sticks into the central pixel, and a variance sensitive conductance functions is designed to adaptively control the diffusion strength in varying directions. The proposed method is in essence an integration of the linear boundary detection operator, i.e. stick technique, and the nonlinear diffusion model. In homogeneous regions, our method will act as a Gaussian like low pass filter, since the sticks are partially overlapped near the center, which implicitly assigns distance dependent weights to neighboring pixels. In heterogeneous regions, the information is expressed as many structures, which often occur as line boundaries or tube shapes in ultrasonic images, then our approach can encourage smoothing along the sticks falling inside the structures, and penalize blurring along the sticks across edges. The performance of our method is verified in experiments of both synthetic and clinical ultrasonic images. The results show that our method outperforms the existed filtering techniques in term of smoothing homogeneous regions, preserving resolvable features, enhancing weak edges and linear structures.

Information hiding is a new technology which integrates with theories and technologies of many academic and technical subjects. For information hiding, digital media are used as the carrier of the information to be hidden. The carrier conceals secret messages by covering the form of their existence. In this paper, we briefly introduce the definition, basic models and basic characters of information hiding. The application and research trends for information hiding system are concerned. The information hiding technology based on digital image processing is closely related to human vision system. When the messages are having been concealed, the human eyes are due to verify the existence of hiding messages. That is, the status of information coverage depends on the human vision system. It is obvious that the characteristics of human vision system is to be taken advantage. The added secrete information in the digital image should have no any effect onto human eyes. In our research work, an implementation of information hiding technology system which is based on digital image encoding is proposed. First by analyzing knowledge of digital image processing and the model of human vision system, we discussed the algorithm of time domain appending method and the algorithm of substitution of lease significant bit. Secondly, we analyzed theory and algorithms of 2-D discrete wavelet transform and frequency domain algorithm based on discrete wavelet transformation. Carefully design software for information hiding based on digital image using Microsoft Visual C++6.0 is implemented. The communication with hiding messages may use any format of images such as BMP. It is proved to be an effective application.

In this paper, we present a three-dimensional (3D) hyperspectral image compression algorithm based on zeroblock coding and wavelet transforms. An efficient Asymmetric 3D wavelet Transform (AT) based on the lifting technique and packet transform is used to reduce redundancies in both the spectral and spatial dimensions. The implementation via 3D integer lifting scheme allows to map integer-to-integer values, enabling lossy and lossless decompression from the same bit stream. To encode these coefficients after Asymmetric 3D wavelet transform, a modified 3DSPECK algorithm - Asymmetric Transform 3D Set Partitioning Embedded bloCK (AT-3DSPECK) is proposed. According to the distribution of energy of the transformed coefficients, the 3DSPECK's 3D set partitioning block algorithm and the 3D octave band partitioning scheme are efficiently combined in the proposed AT-3DSPECK algorithm. Several AVIRIS images are used to evaluate the compression performance. Compared with the JPEG2000, AT-3DSPIHT and 3DSPECK lossless compression techniques, the AT-3DSPECK achieves the best lossless performance. In lossy mode, the AT-3DSPECK algorithm outperforms AT-3DSPIHT and 3DSPECK at all rates. Besides the high compression performance, AT-3DSPECK supports progressive transmission. Clearly, the proposed AT-3DSPECK algorithm is a better candidate than several conventional methods.

Simulation of particle systems is time consuming. However many particle system applications require fast interactive animations. For example, simulation of physically realistic complex snow phenomena is very useful in training, education, and entertainment. In this paper, based on particles system and aerodynamics, we use computational, and behavioral simulation techniques to simulate snow behavior in a flight simulator in real time. At first, we have introduced the principle of particle. The second, this paper has discussed basic attributes of snow particle in detail, we further simplify the numerical calculations according to practically requires and give some new attribute to snow that use a circle shape and specific initializing position of snow. The third, we have given the basic steps to generate a particle system. The fourth, in order to strengthen the depth of field, we make the particle from several kinds areas that faced to our point of view. At last, we construct a new wind models to generate particles and control the behavior accordingly. The approach has been implemented and applied in a flight simulator.

Because of clear image, large angle of view and strong depth of field etc. advantages, the film negative press off-axis type virtual image display system has been used to flight simulator widely. At first, this paper has introduced the frame and principle of film negative press off-axis type virtual image display system in detail, the secondly, we have combine our practice works, deduced calculated formula about the face type of back project screen of film negative press off-axis type virtual image display system by using method of changing light axis. The third, this paper has discussed the key designing and making steps of sphere reflector, at the same time we have given the automatic control method of negative press of reflection mirror. At last, we have given the performance parameters of whole system. It was proved that the film negative press off-axis type virtual image display system have meet all requests of the visual scene of flight simulator by testing, and this system has such virtues that the visual point of the observer may move out of the light axis. Now it has been applied on the visual scene display system of the high performance simulator, the effect is well.

It is very important to understand size and shape of PM2.5 (particulate matter smaller than 2.5μm in aerodynamic diameter) for further studying composition and characteristics of particulate matter in atmosphere. The PM2.5 was collected using no-filter membrane method. The PM2.5 images were observed by Differential Interference Contrast (DIC) Microscopy with 12 million pixels digital static camera, and the images with fine resolution 300nm were obtained. Some images of particles assorted by physical method were presented. In order to count and measure the particles, these images were processed by a series of methods, such as color channel extracting, contrast enhancing, background flattening, Gauss filtering, and scaling with standard objective micrometer, on the Image-Pro Plus (IPP) software platform. The design considerations of parameters in analyzing process were also discussed in detail. The statistical data of particle size and number of PM2.5 were obtained. Results show that the total number of particles <2.5μm in this measurement is made up of 97%, and 29.13% is the particles <1μm.

High efficiency and high automation are essential in the process of metal cutting. How to control the chip will affect processing quality, cutting tool life and productivity greatly. With the development of image processing technology, machine vision has been widely used in real-time monitoring of chip shape. A set of machine vision detection system is developed for realizing image capture, image processing, image pattern matching and image analysis in real time in this paper. Especially, dynamic template is designed to match the complex chip. In this system, LabVIEW is used as system platform, QP 300 picture capture card of Daheng-Image cooperation is used as image capture hardware, LED of CCS cooperation is used as light source. The actual operation shows that this system can identify typical C shape chip and spiral shape chip. Meanwhile, other functions are developed, such as parameter optimization and network transmission.

The goal of structured light techniques is to measure the shape of three-dimensional object using automatic non-contact techniques and based on trigonometry measurement. This thesis deals with the device of obtaining range image based on multi-line structured light. The projector is taken as the projecting source of this device and can project multi-line structured light. The digital camera is taken to obtain stripe image and can capture the stripe image in a single frame. The technique of space encoding is utilized to process the stripe image and could avoid the aliasing of the multi-line structured light. This technique can ensure the veracity of the range image obtaining. The experiments have shown that the range image obtained by this device have better results and higher speed of acquisition. Measuring range of this device is 200mm(X)×120.0mm(Y)×150.0mm(Z). The measuring error of Y and Z direction is less than 0.3mm. The measuring error of X direction is less than 0.1mm.The absolute error of the measurement result in the direction of Z is less than 0.2%.

In order to utilize the multiscale characteristics of wavelet transform more sufficiently, a new adaptive image restoration method using wavelet packets based on the edge detection is presented. The method allows for the decomposition of image signal with various frequencies in the subband domain. The proposed filters in the paper explicitly incorporate both within and between subband relations of the decomposed image. The adaptive filter is used in the approach for considering local adaptation in each subband of the best basis. In order to keep the high frequency information of the image, the adaptive edge detection is presented in the paper. Experimental results, which could test the proposed method, are got. It is found that the proposed method show not only great noise reduction in the processed images but also significant improvement of subjective image quality over the conventional image restoration methods.

With the applied computer vision technology, the boot of automobile shaft is measured real-timely under a high rotation speed and high temperature condition, and an extraction method of pixel-level object profile curve is introduced. Besides, the radial dilation of the boot is obtained accurately through averaging several measurement results on a basis of pixel-level edge. During the test course, the boot not only dilates in radial direction but also deforms in axial direction. According to the characteristic of the measurement object variation, this paper presents a modification algorithm, which processes departure in segments along the X coordinate, to reduce the measurement deviation from deformation in axial direction, and the measurement accuracy of the boot profile dimension is improved.

Aiming at the problem of process monitoring on chip generating in automatic machining, methods of recognizing chips' shape based on neural net are researched in this paper. The conception of area ratio of the chip image to the located window is defined, the area ratio feature has been proposed because the size of all windows and the direction of chips are respectively same. At the same time, the Euler number characteristic and disperse degree characteristic of the chip image have been worked out. The above geometry characteristics of the chip image are chosen as input vectors of neural network, and the 50 various images of each type such as C shape, spiral shape and disorderly shape are chosen as training sample, the recursion least square law is used to train network. The recognition rate and training time of the BP network are compared with those of the RBF network, so the conclusion that the RBF network is superior to the BP network at the aspect of chip shape recognition has got, and the relevant computer program has been developed, which possess good real-time application and adaptability by way of the experiment certification. The recognition rate achieves more than 90%.

Dental pulp is located in root canal of tooth. To modern root canal therapy, "Root canal preparation" is the main means to debride dental pulp infection. The shape of root canal will be changed after preparation, so, when assessing the preparation instruments and techniques, the root canal shaping ability especially the apical offset is very important factor. In this paper, a novel digital image processing and measurement system is designed and applied to quantitative analysis of simulated canal shape. By image pretreatment, feature extraction, registration and fusion, the variation of the root canals' characteristics (before and after preparation) can be accurately compared and measured, so as to assess the shaping ability of instruments. When the scanning resolution is 1200dpi or higher, the registration and measurement precision of the system can achieve 0.021mm or higher. The performance of the system is tested by a series of simulated root canals and stainless steel K-files.

In 3D model retrieval, feature extraction of 3D model is a very important topic. In this paper, the goal of the proposed approach is to decompose 3D model to meaningful components by iterative scheme and hierarchical fuzzy clustering, then construct a node graph to represent 3D model. The similarity calculation between two 3D models is processed using a coarse-to-fine strategy. Experiment results show that out method is suitable for part matching, articulated matching and global matching of the models.

In this paper, we propose a new method for finding similarity in 3-D protein structure comparison. Different from the other existing methods, our method is grounded in the theory of fractal geometry. The proposed feature vectors of protein structures are invariant to the rotation, translation, scaling of the protein molecule, and it is simple to implement. The method is very fast because it requires neither alignment of the chains nor any chain-chain comparison. We calculate the fractal features of a set of 200 protein structures selected from PDB (Protein Data Bank). The experimental result shows that our method is very effective in classification of 3-D protein structures.

It is very difficult to measure the distance of a moving point target which is not cooperative for detecting in the remote sensing field, because that the point target has no geometrical dimensions and textures can be used and is easy to be missed. In this paper, a new algorithm based on the image sequence and the nonlinear regressive filtering algorithm is proposed, in order to determine the 3-D parameters of the moving point target in an efficient passive way. And a new multi-channel optical imaging system is designed, which is composed of a high- resolution center imaging system and four low-resolution sub-imaging systems. By the geometrical relationship of the four sub-imaging system, the initial values of the nonlinear regressive filtering algorithm for estimating can be obtained easily. Finally, the experiments of the proposed algorithm have been done on a real system, and the results proved that the algorithm could passively obtain the 3-D parameters of the moving point target efficiently. Furthermore, in the estimating procedure, the character of the nonlinear regressive filtering algorithm saves lots of memory units and reduces the computing quantity.

At present, in the field of image processing, the main algorithm to restore the blurred image is the blind deconvolution. But most of the blind deconvolution methods have to iterate a large amount of times and the result is also unsatisfactory. In this paper, a new blind deconvolution algorithm is proposed, which, consisting of two steps, is based on simultaneous estimating the specimen function and the parameters of the point-spread function (PSF). Firstly, it uses the expectation maximization algorithm (EM) to iterate the specimen function; secondly it uses the conjugate gradient method to estimate the parameters of the PSF. The mathematical model ensures that all the constraints of the PSF are satisfied, and the maximum-likelihood approach ensures that the specimen is nonnegative. In this paper, the general Gauss function is used to be as the PSF. In the experiment, it can successfully restore both the two-dimensional and three-dimensional images within limited times of iteration.

An algorithm for maximum-likelihood image restoration based on the expectation maximization (EM) algorithm is proposed in this paper. This estimation is based on a depth-variant imaging model in three-dimensional optical sectioning microscopy. As a result of the refractive index mismatch between the immersion medium and the mounting medium of the specimen, the imaging model in three-dimensional optical-sectioning microscopy incorporates spherical aberration that worsens with increasing depth under the coverslip and changes in the point spread function (PSF). Two-dimension images restoration and three-dimension serial images restoration are to be used to analyze the capability of the EM-ML algorithm, and the performance shows that the EM-ML algorithm can restore the blurred of image by the depth variant image model.

Resole and novolac are the coating resins of the CTP (Computer to Plate) thermal cross-linking plate and their characters directly determine the light sensitivity of the plate. In this paper, these resins were studied, and experiments showed that when the novolac's molecule weight is about 5000, the light sensitivity and the quality of the plate are the best. The optimal conditions of synthesizing resole were obtained, and more experiments were attempted to improve the light sensitivity of the plate through altering the proportion between the two image resins.

The relative motion to the CCD camera during the imaging course is becoming more and more prevalent. Especially when the objects are moving forwards or backwards at a high speed, the image's resolving power will decline rapidly from the center to the edge of each frame, which will deteriorate the imaging quality seriously. On the problem, in view of the use of grating with sine distributed intensity in the evaluation of static system's imaging quality, and considering that during the objects' forward-reverse motion, the magnification of optical image of the object is changing smoothly, a model of concentric circles with sine distributed intensity is built on the area CCD's photosensitive surface, the contrast of which is 100%. Using the concentric circles model, the paper has deduced the signal intensity and contrast expressions of each frame image during the objects' relative motion. Then the CCD detector's imaging quality can be evaluated by the signal intensity and contrast of each frame image. From the expressions, we can know that the concentric circle's radius, the object's spatial frequency's change rate and CCD's integration time will all influence the image's resolving power. Whichever parameter of the three increases, the contrast will decline by the function of |sinc|. On the basis of the theoretic analyse, the paper has given some pictures and graphs of the experimentation. The calculating data and pictures from the experimentation have verified the results of the theoretic analyse.

Vector Quantization is one of the most powerful techniques used for speech and image compression at medium to low bit rates. Frequency Sensitive Competitive Learning algorithm (FSCL) is particularly effective for adaptive vector quantization in image compression systems. This paper presents a compression scheme for grayscale still images, by using this FSCL method. In this paper, we have generated a codebook by using five training images and this codebook is then used to decode two encoded test images. Both SNR and PSNR and certainly the visual quality of the test images that we have achieved are found better as compared to other existing methods.

A method for the detection and tracking of human face in color images is described in this paper. A skin color model in r,g,b chrominance space is used for segmenting skin color regions from non-skin color region. The best-fit rectangle of the skin color region is labeled as a candidate face. But the face skin color is sensitive to the change of the environment illumination. An algorithm is proposed to update the skin color model's parameters in time so that the model is adapted to different lighting conditions. The confidence measure is presented to evaluate the reliability of skin color model. Experiments demonstrate that the self-adaptive color model is more effective than the fixed model. The color adaptation makes that the color model can be better fit to the more complex application environment.

New structure low intensity x-ray image system is mainly made of plane plate mode x-ray intensifier of single proximity focus and CCD data acquisition and processing system. The paper explains the noise source and characteristic of the low intensity x-ray image system. By the system composition, the image noise source of low x-ray imaging system is constituted with quantum noise, particulate noise and dark noise of CCD. Then the compound methods of the "multi-frame mean + morphological transform filter" is submitted which deals with the imaging noise. Firstly, some frame images is superimposed, then mean image is calculated from those images, which is under the principle of noise non-correlation. Secondly, distinguishing with the conventional ways of morphological transformation filtering algorithm, the differential image information is referred to de-noising. Under the multi-scale morphological thought, the differential image which is obtained from the source image includes noise and some image details. After the noise of the differential image is cut off by the wavelet translation, the differential image is added to the last filtered image by the multi-scale morphological filter, then the clean image is achieved which has no noise but keeps the image details.

An improved chaotic algorithm for image encryption on the basis of conventional chaotic encryption algorithm is proposed. Two keys are presented in our technique. One is called private key, which is fixed and protected in the system. The other is named assistant key, which is public and transferred with the encrypted image together. For different original image, different assistant key should be chosen so that one could get different encrypted key. The updated encryption algorithm not only can resist a known-plaintext attack, but also offers an effective solution for key management. The analyses and the computer simulations show that the security is improved greatly, and can be easily realized with hardware.

Hyperspectral image can be analyzed by Convex Geometry Analysis(CGA) method. CGA method can unmix endmembers from hyperspectral image. The endmember proportions of mixed pixels can be calculated in inherent dimensional space, and a proportion image, which is called inherent proportion image, is obtained. The endmember proportions of mixed pixels can be calculated in n-space by the Constrained Least Squares, and a proportion image, which is called CLS proportion image, is obtained. In this paper, the inherent proportion image and CLS proportion image of a 30-band remote sensing image are obtained. The two proportion images are similar. The targets that are smaller than earth surface pixel can be identified by the inherent proportion image.

We analyze the inherent channels of hyperspectral data with convex geometry analysis method. In this paper, a method-Volume Method, which selects the inherent channels of hyperspectral data, is presented. The concept of convexity geometry can be used to great advantage in the analysis of hyperspectral data. Convex simplex and inherent dimensionality concept is discussed on base of convex geometry. A set of 252-band hyperspectral data is applied to testify the Volume Method. The endmember proportions are calculated in the inherent dimensional space whose channels are selected by the Volume Method, compared with Constrained Least Squares Method in 252-space.

In this article, we proposed an effective adaptive 2-dimension blind watermarking algorithm based on feature of a remote sensing image. This algorithm exploited a gray image as the watermark, pretreated the watermark image by Arnold confusion and wavelet compression, and embedded it into the selected subband of wavelet transformation domain of the remote sensing image according to neighboring symbol's mean value and odd-even adjugement rule, moreover, detected watermarks without the original remote sensing image. The attack analysis and experimental results show that the watermarking algorithm is transparent and robust, with accurate watermark detecting results and low complexity, and it also has strong robustness against various image attacks such as JPEG lossy compression, median filtering, additive noise, scaling, cropping, rotation, random geometrical attack and Stirmark attack. Furthermore, after embedding watermarks, there is almost no influence on such applications of the remote sensing image as edge detection and image classification.

Recently, wavelet transform has been applied widely in watermarking research as its excellent multi-resolution analysis properties. Almost all previous works on watermarking capacity are realized in spatial domain. This paper proposes a new adaptive watermarking capacity analysis method in wavelet domain. This paper also analyzes the relationship between watermarking capacity and watermarking detection bit error rate (BER), and derives the relation between capacity and the limit of BER. According to the result of research, watermarking detection BER is mainly influenced by the watermarking average energy and watermarking capacity. The BER rises with the increase of watermarking capacity.

With rapid development of electronic imaging and multimedia technology, the telemedicine is applied to modern medical servings in the hospital. Digital medical image is characterized by high resolution, high precision and vast data. The optimized compression algorithm can alleviate restriction in the transmission speed and data storage. This paper describes the characteristics of human vision system based on the physiology structure, and analyses the characteristics of medical image in the telemedicine, then it brings forward an optimized compression algorithm based on wavelet zerotree. After the image is smoothed, it is decomposed with the haar filters. Then the wavelet coefficients are quantified adaptively. Therefore, we can maximize efficiency of compression and achieve better subjective visual image. This algorithm can be applied to image transmission in the telemedicine. In the end, we examined the feasibility of this algorithm with an image transmission experiment in the network.

The television imaging system of steady platform is designed by applying optical zoom lens principle and array CCD image collecting technology in this paper. The system is a crucial component of steady platform to come true tracking capability. The paper expatiates working principle and its key technology.

Pattern formation is a process by which a spatially uniform state loses stability to a non-uniform state. A usual effective method for studying the property of patterns is to take pictures or record images of the patterns with digital cameras and get useful information from them. In this work, the structure, the spatial frequency spectrum of the patterns, the precise location of the discharge filaments center and the distribution of the light emission density of filaments are analyzed with Matlab6.1. This work can provide a beneficial reference for researchers who study pattern dynamics in DBD system or other systems.

Traditionally, the grade discrimination and classifying of bowlders (emeralds) are implemented by using methods based on people's experiences. In our previous works, a method based on NCS(Natural Color System) color system and sRGB color space conversion is employed for a coarse grade classification of emeralds. However, it is well known that the color match of two colors is not a true "match" unless their spectra are the same. Because metameric colors can not be differentiated by a three channel(RGB) camera, a multispectral camera(MSC) is used as image capturing device in this paper. It consists of a trichromatic digital camera and a set of wide-band filters. The spectra are obtained by measuring a series of natural bowlders(emeralds) samples. Principal component analysis(PCA) method is employed to get some spectral eigenvectors. During the fine classification, the color difference and RMS of spectrum difference between estimated and original spectra are used as criterion. It has been shown that 6 eigenvectors are enough to reconstruct reflection spectra of the testing samples.

On the basis of spectrophotometric color matching, the color difference weight factor is proposed and used in the thesis. The weight factor can be expressed as ω_j = {{[(x(λ_j)]² + [y(λ_j)]² + [z(λ_j)]²}[S(λ_j)]²}^½ and obtained according to the assumption of Σ_j(ΔX_j)²+(ΔY_j)²+(ΔZ_j)² -> min, i. e., in the range of visible spectrum it is assumed that the square sum of tri-stimulus value deviation produced by spectrum deviation at each wavelength is minimal.Through comparison with spectrophotometric color matching, we find a new weight factor. The new factor multiplied by the variety of reflectivity is the color difference, which is cause by the difference of reflectivity between standard color and matching color. So we name the weight factor: color difference weight factor. The prediction of computer shows the color difference produced by the weight factor is smaller than that produced by the two weight factors which were designed by Schmid and Strockash.

View range is one of the main integrated performance indexes, and is also the important performance index needed to give when night view systems leave factory. A new objective view range evaluation method for night view system is proposed in this paper by using image processing technology, data fit, vision sense formula under glimmer light circumstance and other knowledge. With the help of the method, the paper has carried out four groups' experimentation with field situation of four kind's night view system. The result of view range calculation indicates that the new method not only has effectively improved the speed of view range measurement for night view system, but also possesses good precision and practicability.

Based on analysis of the working principle of ICCD imaging system, the paper develops a new testing system for synthetic performance of ICCD imaging system. With the help of this testing system for integrative performance and the theory of Weibull psychometric function, a new method for MRC measurement of Space Light Modulation (SLM) is proposed, and the paper firstly implements measurement for minimum resolvable contrast performance of ICCD imaging system. The method provides reference for MRC performance measurement of other imaging system.