最新刊期
卷 7 ， 期 11 ， 2002
- 摘要：The problem of restoring degraded images, and blind images especially, arouses many scholars' interest, since the prospects of their applications are wide. Neural network with many inherent merits is attached importance by researcher in all fields. Its application on images processing comes in for great attention. The development trend is transiting from semi blind image restoration to blind images restoration. Firstly, this paper provides simply a background for reader regarding the subject of image restoration. Understanding the fundamental theory, methods and concepts, and the current state of the art in the field of image restoration is vitally important to the understanding of the image restoration using neural network. Secondly, this paper presents a short description on image restoration using neural network and its mathematical model. Finally, we give mostly an overview of neural network for image restoration; including the background, current algorithm, applications and research topics, in the effort to motivate the research works on this field in our country.2518|374|0更新时间：2024-05-07
- 摘要：This paper proposes a hierarchical method of image segmentation with topological adaptability, called B spline active contour based on edge detector. Unlike traditional method of active contour, our method takes region homogenous property into account and designs a new external force regional force, which is very robust to noise contamination. Also, internal force is integrated into the B spline. Our method is composed of two steps. First step is a kind of low level image segmentation. In this step, a local edge detector is used for detecting all edge points of the image. Second step is a kind of high level image segmentation. In this step, our B spline active contour based on the global image statistic is used for refining the region boundary. Also, we propose a new topology adaptability method, which is based on the change of region Euler number. Our method requires less interactive operation and is insensitive to initial condition. The experiments reported in the paper, performed on real images, confirm that the method can offer a good segmentation result and it has a very good topological adaptability.2296|334|0更新时间：2024-05-07
- 摘要：The immune cell image must be exactly segmented first in order to realize cell's parameter measurement and get a right analysis conclusion. In this paper, an effective immune cell image segmentation algorithm based on mathematical morphology is presented. In order to get better segmentation results in addition to the morphology based watershed growth algorithm the histogram potential information is involved, which means, the image spectral information is combined with spacial information. How to get the exact segmentation result is a major issue for immune cell analysis. Watershed growth combines the basic idea of region growth and edge detection and has the advantages of both the method. Using the method, single pixel width, connected and closed object boundary can be detected automatically, which is necessary for cell image segmentation. But obtaining an effective and credible marker is a crucial step of watershed segmentation. By involving the histogram potential function, the markers suitable for watershed segmentation can be clearly improved. By this method, a segmentation result quite consistent with human vision can be gotten, and both the segmentation speed and repeatability meet the medical clinic need, and the analysis conclusion accords with clinic diagnoses.2694|271|0更新时间：2024-05-07
- 摘要：A fast moving object segment algorithm on moving video sequences is proposed. Firstly, the difference image is produced between two continuous frames which are filtered in advance by Gauss template. Then the author proposed "assimilation" filling technique and a template modification method employing dominant motion vector are applied to remove the uncovered background and get the final object template. The "assimilation" filling technique only fills the abruptly changed gap on the edge or the inner hole, while leaves the smoothing changed edge untouched. What need to do in the filling process is just to increase or decrease the coordinate, not involving multiplication and division operation, so it is very fast. A memory buffer of objects is maintained through the whole process to remedy the disfigurement of regions gotten from a single frame. The algorithm doesn't rely on a fixed background, and can remove uncovered background to get the accurate size of objects. The algorithm only uses two continuous frames to segment the moving object from background, not like the symmetrical difference based methods which make use of three frames. It is a simple, but fast and robust algorithm.2584|425|0更新时间：2024-05-07
- 摘要：Video segmentation is an important step in image recognition and in video coding based on MPEG 4 standard. This paper proposes a robust video segmentation algorithm which can greatly segment not only multiple VOs but also single VO from the static background. From the research, it is found that the raw contours of the active objects in the sequential video frames can be gained fast and exactly by using symmetrical DFD. So first use every three continuous video frames to get symmetrical DFD, then acquire the motion area of the moving object. Second analyze the result of the symmetrical DFD by clustering to inspect the number of VOs in the video frame. During clustering analyzing, not employing the traditional fuzzy C clustering mode, but employing a new clustering mode with using the body's characteristic which is proposed firstly in this paper. Third use the active contour algorithm in the binary image to get the contours that enclose the VOs accurately. By this step acquire the result of segmentation. Last, track the VOs in the video frames by the vectors of optical flow and also modify the result in some special conditions. All tests receive the ideal segmentation effects by using this algorithm with several image sequences. The experimental results show that this algorithm can complete video segmentation both for multiple VOs and for single VO in the static background correctly, so this algorithm has some robusticity and practicality.3302|385|0更新时间：2024-05-07
- 摘要：Caption text presented in the video plays an important role in video retrieval and browser as it provides highly condensed information about the contents of the video. The caption text only can be used after it is extracted from video. According to Morrone's phase congruency theory, image features such as edges, shadows and bars always occur at points of maximum phase congruency, and the maxima of local energy occur at points of maximum phase congruency. Based on this theory, a video caption text segmentation approach is presented in this paper. Instead of Morrone's approach in calculating of local energy, we extended Morrone's approach through constructing a quadrature pair filter from a biorthogonal wavelet by Hilbert transform. The local energy is then calculated from the multiresolution decomposed image in octave bands. According the relationship between local energy and caption text's edge features, we then segment the image using local energy projection. This is the first step of caption segmentation. After that, colour segmentation can be applied to the first segmentation results. The experiments in this paper show that this approach can achieve good caption region segmentation results.2610|348|0更新时间：2024-05-07
- 摘要：The difficulty of a face recognition problem is to handle different types of variations, such as facial expression, illumination and pose. In order to improve the robustness of face recognition with respect to facial expression, this paper proposes a new approach, the eigenmotion based method, which is tolerant to large variations of facial expressions. In this new approach, first motion vectors are computed between a testing face image and a neutral training image using the block matching method, then projected to a low dimensional subspace that is pre trained by applying principal component analysis(PCA) to motion vectors resulting from training images with expression variations. This subspace is called an eigenmotion space. Finally the identification of the testing image is determined based on its residue to the eigenmotion space. Both the individual modeling method and the common modeling method are described in this paper. Experimental results show that the proposed eigenmotion based method outperforms the eigenface approach in the presence of facial expression variations. The approach can be extended to model other types of variations as well, for example, illumination and pose variations.2830|386|0更新时间：2024-05-07
- 摘要：With the development of the advanced techniques of human computer interaction(HCI), gesture recognition is becoming one of the key techniques of HCI. Due to some notable advantages of vision based gesture recognition(VGR), e.g. more naturalness to HCI, now VGR is an active research topic in the fields of image processing, pattern recognition, computer vision and others. The method of model matching using Hausdorff distance has the characters of low computing cost and strong adaptability. The system described in this paper applies the hausdorff distance for the first time to visually recognize the chinese finger alphabet(CFA) gestures(total 30 gestures) with the recognition features of edge pixels in the distance transform space. In order to improve the robust performance of the system, the modified hausdorff distance(MHD) has been proposed and applied in the recognition process. The average recognition rate of the system using MHD is up to 96 7% on the testing set. The experimental result of the system shows that using the method of model matching based on the Hausdorff distance to realize the vision based static gesture recognition is feasible.4137|368|0更新时间：2024-05-07
- 摘要：Support vector machine(SVM) is a novel type of learning machine, this thesis introduces the theory of SVM briefly and application in a classification system for texture image, and discusses in detail the core techniques and algorithms, which combine SVM and distance classification into two layer serial classifier. SVM has shown to provide better generalization performance than traditional techniques. However, because using Quadratic Programming (QP) optimization techniques, the training of SVM is time consuming, especially when the training data set is very large. So we have two classifiers combined. Firstly, a rejecting coefficient and rejecting rule are defined. According the rejecting rule, the distance classifier can classify the images and give the final results, or reject to classify the input images. The rejected images are fed into SVM for further classification. The algorithms can take advantages of SVM and distance classification. The experiments show that the algorithms have low error rate and high speed.2470|390|0更新时间：2024-05-07
- 摘要：Image classification system is an important part of any information retrieval system and pattern recognition system, and its key issue is to select some appropriate feature bindings of an image. Recent years content based image retrieval has been a very active research area. The dimension of the image feature vectors is normally very high and it's hard to index images. One of the main challenges in content based image retrieval is to develop techniques of performing dimension reduction. In this paper, a new model of searching multiple classification criterions has been proposed in which different feature bindings were formed to find new classification criterions, and a new algorithm was designed for this model. The experimental results shown that the proposed model can perform dimension reduction. The algorithm for the model is capable to reduce computational time which was also illustrated with results. The multiple criterions in combination with the information retrieval techniques can implement personalized information retrieval, and some results were given in last section.2308|317|0更新时间：2024-05-07
- 摘要：Iterated function system(IFS) is an effective method to define and describe fractals. An IFS determines only one fractal which is called attractor. Although random iterated algorithm proposed by Barnsley can display easily and quickly an attractor of IFS on computer screen, it is not sure to generate all points of an attractor within any limited steps. In order to overcome the shortcoming of the algorithm, a new algorithm of gradually computing IFS attractor from one fixed point of an invertible affine transformation is presented. Because of self_similarity of an attractor of IFS there exists similarity between different regions of an IFS attractor, With this property different parts of an attractor can be showed one by one. The experimental results prove this method is feasible. A whole attractor can be computed through limited steps by using this algorithm, and unlike random iterated algorithm probability is not necessary.2978|305|0更新时间：2024-05-07
- 摘要：In this paper, a new 3D soft tissue display scheme is proposed, which consists of four steps: image segmentation, distance transformation, peeling operation and volume rendering. Firstly, an image segmentation method is adopted to detect the contour of skin, and a new binary output image is produced. Secondly, to reduce the computation time, a new 3D Euclidean distance transformation algorithm was adopted to compute the distance map of the medical image. Thirdly, the image data of the scarfskin and subcutaneous fat of the specified depth is removed through peeling operation. Finally, in order to meet the requirement of real time interaction in the medical application, a new multi surface visualization method was adopted, whose rendering time is reduced by only projecting the voxels near the boundaries between the different tissues. Meanwhile, it improves the visualization quality by computing the normal of the point where the ray crosses the iso surface in the projected voxel. This scheme was implemented in the 3D medical image process and analysis system developed by our lab to display the soft tissue of the CT images. The experiment results show that the anatomical structure among the blood vessel, flesh and bone can be visible clearly, and this method has important practical value in clinical diagnosis.2513|333|0更新时间：2024-05-07
- 摘要：During the process of target tracking, the template should be updated properly in order to adapt the change of the target pose. In this paper, a new template update method is proposed. On the base of affine tracking, moving object detection is performed. We estimate the pose change of the object according to the detection result image. A new update method is used as follows: Firstly, pixels belong to target and those belong to background are updated with different power; Second, because the template size does not change during tracking, some strategies must be adopted: if the object is turning bigger, the old template is stretched to the proper size and placed in the new template's proper position, the rest pixels in the template are determined by the corresponding pixels in the input image. Otherwise, if the object is turning smaller, the old template is shrunk to the proper size and placed in the new template's proper position. The rest pixels in the template are determined by the corresponding pixels in the input image in the same way. Experiments using these algorithms show this method could keep robustness under complicated environments.2613|323|0更新时间：2024-05-07
- 摘要：In radiosity and other global Illumination algorithm, visibility test is the key component, which dominates the performance of algorithm. It consumes much time and memory. So our goal is to reduce the time spent in visibility test and improve the performance of the shadow. This paper introduces an optimized algorithm to reach this goal. The most popular method to test visibility is ray casting. Based on the theory of ray casting, two ways are used to improve the performance. First, the Shaft culling algorithm was introduced to improve the efficiency of ray casting. It preprocesses to cut off a majority of objects that need not to be test. This algorithm is very useful for decreasing the computation of ray casting. Second, do visibility based meshing. It makes a difference between full visibility, full occlusion and partial occlusion. This guarantees partial occlusion is almost eliminated. This work not only decreases the computation cost and error, but also ensures the quality of the shadow boundary. The two ways are used in the implementation of HRC. Results show the new method is more efficient and the shadow created is very smooth.2651|383|0更新时间：2024-05-07
- 摘要：Considering the object based video standard MPEG 4 , this paper proposes a new motion estimation method--pseudo diamond searching algorithm(PDS), which uses motion vector as referent variable to execute a polygon matching procedure. The motion vectors in near blocks and points in same blocks between sequent frames are correlated. Considering these characteristics, pseudo diamond searching method uses a direction selective strategy to reduce the average searching points and enlarge the searching region. This strategy is efficient under the conditions where movement is complex and vehement. Using C programming and using matlab kits to test an image sequence, this paper shows that such method, compared with the diamond search method(DS) and Hexagon based Search Method (HS), keeps the image quality efficiently as well as reducing the computation complexity.2947|379|0更新时间：2024-05-07
- 摘要：This paper presents a trade mark retrieval method in which the shape feature and spatial relationship are both used for the purpose of making full use of image info and improving retrieval precision. Since trade mark is artificial image and it usually consists of several geometric figures which have sharp edges, so we can look on trade mark as a combination of such regions. For these combinations, invariant moment is used to measure the shape similarity between two trade marks firstly and then the spatial relationship inside the combinations is matched by the way of project clustering. The method mentioned in the paper is just based on the idea, considering both the shape feature and spatial relationship of each part of trade mark at the same time. The whole retrieval process is then divided in two stages, i.e., rough retrieve and refine retrieve. Since we take the two factors of trade mark into account simultaneously, so we can ensure the consistency in both the whole and local sides. Compared with the way of only using shape feature to retrieve images, the results of experiment show this method has higher precision and the output accords with people's visual feeling better.关键词：Trade mark;Shape correspondence;Spatial position;Image retrieval2755|383|0更新时间：2024-05-07
- 摘要：In laser scanning confocal microscopy imaging, CT imaging, MRI imaging, and in 3D image processing and recognition, visualization of 3D objects is essential. Nowadays, the popular ISO surface visualization is based on 3D virtual object reconstruction. The 3D virtual object reconstruction can be volume based. It can also be 2D slice based. In the 2D slice based, due to the still existing problems, such as contour pairing, surface diverging, contour pair patching, etc, a few new methods are proposed for the problem solving. In theses methods, the contour paring is carried out with OR and AND operations. For two contours, if the results of the OR and AND operations satisfy a pre set criterion, the two contours are a pair. The surface diverging is decomposed with mathematical morphology. The boundaries generated by the morphologic operation are that of the diverging surfaces. The contour pair is patched with triangles to form the ISO surface after the pair is polygonized. The triangles are constructed under the criterion of minimum contour linking edge. With the methods, some experiments have been done. The experimental results showed that the theories of the methods coincide perfectly with the practice. The advantages of the methods are simple programming and fast computation.2599|372|0更新时间：2024-05-07
- 摘要：Due to the enormous magnitude and unstructured contents of multimedia data, solutions must be provided for their effective compression and efficient indexing, in order to realize all kinds of multimedia applications. However, the traditional approaches treat compression and indexing problems separately during the past decades. The compression algorithms are implemented without indexing supported in compressed domain while the indexing operations are mainly undertaken in original format of multimedia data, resulting in lower overall performance of current multimedia application system. In order to improve the situation, a joint image compression indexing algorithm based on iterative fractal method is proposed in this paper. Firstly, the iterative fractal method is employed to compress the image in wavelet domain for effective compression. Then feature vectors representing the distribution properties of IFS(Iterative Function System) are constructed to support the indexing of images based on the fractal coded image data. Simulation results verify the efficiency of the methods and show the potentials of the fractal based image indexing methods.2600|326|0更新时间：2024-05-07
- 摘要：When the compressed bitstream generated by a H.263 video coder is transmitted over the Internet, it is sensitive to packet loss. Due to the coding structure of the H.263 video compression algorithm, the packet loss can effect not only the current picture frame but also successive frame so that the quality of the video will be degraded seriously. The error concealment can be used to eliminate the effect. It commonly includes two methods : spatial concealment and temporal concealment. Spatial concealment uses the nearby correct image data of the corrupted part to recovery the image quality.The latter depends on the image data of the previous frame to estimate the lost data.However, spatial concealment will cause the blur of image quality and simple temporal concealment will make obvious artifacts. In this paper, we propose a temporal concealment algorithm based on block match principle. And in order to reduce the computational complexity, a new three step search algorithm is substituted for the exhaust search. Simulation results show that the proposed algorithm can produce acceptable quality of picture and can be implemented in the video conference application.2381|300|0更新时间：2024-05-07
- 摘要：Algorithms to reconstruct digital elevation map(DEM) from elevation contour is widely used in many applications. In the process of designing a 3D terrain visualization program, where only elevation contour map is available, a new method, region intra interpolation algorithm, was proposed. It utilizes the structural features of elevation contour images, that is, image is partitioned into many regions by elevation contours and each region is only bound to at most 2 elevation values. Therefore each point inside this region can be calculated as a linear interpolation from the two boundary values. The problem then turns into finding the two shortest distances to its bounding elevation contours. Fast realization is also given out, and this speed up also achieved by special design utilizing elevation contour structures to save time on searching for the closest points. Compare with current algorithm like Quadrant Search algorithm, which is design based on a model of random distributing of known value, it shows good performance on both speed and accuracy.2766|329|0更新时间：2024-05-07
- 摘要：With the development of the advanced techniques of human-computer interaction (HCI), gesture recognition is becoming one of the key techniques of HCI. Due to some notable advantages of vision-based gesture recognition(VGR), e.g. more naturalness to HCI, now VGR is an active research topic in the fields of image processing, pattern recognition, computer vision and others. The method of model matching using Hausdorff distance has the characters of low computing cost and strong adaptability. The system described in this paper applies the hausdorff distance for the first time to visually recognize the chinese finger alphabet(CFA) gestures (total 30 gestures) with the recognition features of edge pixels in the distance transform space. In order to improve the robust performance of the system, the modified hausdorff distance(MHD) has been proposed and applied in the recognition process. The average recognition rate of the system using MHD is up to 96.7% on the testing set. The experimental result of the system shows that using the method of model matching based on the Hausdorff distance to realize the vision-based static gesture recognition is feasible.2561|314|0更新时间：2024-05-07
- 摘要：Multisensor data fusion technique has been widely used in remote sensing image processing. A major problem in using remote sensing multispectral images is the low spatial resolution. In order to improve spatial resolution of the multispectral images, a multiresolution image fusion method based on the normalized correlation moment is proposed in this paper. After 2 D wavelet transform the normalized correlation moment is defined by means of statistic first order and second order moment of those wavelet coefficients with high frequency component under different resolutions. By using the local correlation moment as measure of feature selection, the new method carries out multiresolution images fusion. And the fused images include more useful information for further application. Experiments are conducted on Landsat TM and SPOT data. Computer simulations show that the fused images obtain better results in terms of both preserving spectral information and improving spatial resolution of the multispectral images, and are rather used to many aspects of remote sensing image processing such as target recognition, ground materials classification, etc.3099|390|0更新时间：2024-05-07
- Postal code：100190
- Tel：010-58887035/58887030/58887418 Email：jig@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备05080539号-4
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0