最新刊期
卷 8 ， 期 5 ， 2003
- 摘要：This is the eighth in the survey series of the yearly bibliographies on image engineering in China. The purpose of this survey work is mainly to capture the up-to-date development of image engineering in China, to provide a convenient means of literature searching facility for readers working in related areas, and to supply a useful reference for the editors of journals and potential authors of papers. Considering the wide distribution of related publications in China, 545 image engineering research and technique references are selected carefully from 2426 research papers published in a set of 15 Chinese journals. These 15 journals are considered as important journals in which papers concerning image engineering have higher quality and are relatively concentrated. Those selected references are classified first into 5 categories (image processing, image analysis, image understanding, technique application and survey), and then into 21 classes according to their main contents. Some analysis and discussions about the statistics made on the classification results are also presented. This work shows a general and off-the-shelf picture of the various progresses of image engineering in China. In 2002, the number of research papers in image engineering has a considerable increase. Except "traditional" image segmentation and image coding, new research areas, such as image formation techniques, digital image watermarking, human face and organ detection, image matching and information fusion, and image and video retrieval are still in fast progresses.3532|222|0更新时间：2024-05-07
- 摘要：The advent of compression standards, such as JPEG and MPEG, has led to the popularity of the compressed form of image data, and that has brought on the proliferation of image retrieval techniques in the compressed domain. In this paper, we make a comprehensive review and discussions on the compressed domain retrieval techniques proposed in the literature, including the lastest achievements in this field. First, we give the general concepts of the image retrieval technology. Secondly, we analyze different retrieval techniques, including transform domain techniques using Fourier transform, discrete cosine transform, subbands and wavelets, and spatial domain techniques using vector quantization and fractals. Thirdly, we discuss and compare these image retrieval techniques and draw some useful conclusions. In addition, an example application of image retrieval in the compressed domain is presented. Finally, we make a discussion on some open problems and point out possible directions for further research.关键词：Computer image processing;Image retrieval techniques;compressed domain;Discrete cosine transform(DCT);wavelet;vector quantization;fractal2424|190|0更新时间：2024-05-07
- 摘要：In this paper, a novel genetic algorithms with sexual reproduction is proposed to combat premature convergence inherent in Standard Genetic Algorithms(SGA) and speed up convergence. It imitates the sexual reproduction that is very popular in nature: (1) Each individual is encoded using diploid chromosomes which can save more information so as to memorize more good patterns, (2) There is a pair of sexual chromosome that reflects the sexual feature of each individual, so there are two kinds of individuals—male and female individuals, (3) During the reproduction procedure, each individual can only be matched with another individual with different sexual feature, and (4) Dominant genes decide the individual characters. Also, the corresponding crossover, mutation and selection operators for the sexual reproduction are developed in this paper. In the evolutionary procedure, the male individuals reserve higher mutation rate to obtain better global exploring ability while the female individuals have lower mutation rate to enhance local searching ability. As a result, the male individuals possess strong global exploring ability and the female individuals possess strong local searching ability. At the same time, the diploid encoding and dominance law diversify the gene pool. So the algorithm can help the evolutionary procedure to escape from possible local entrapment and obtain good tradeoff between exploration ability and exploitation ability. The experiments are taken on two types of optimization problems, (1) find maximum of minimum values of a series of classical and typical complex multi-modal functions, and (2) find the optimized rout for TSP problem. The experimental results have shown the good performance of genetic algorithms with sexual reproduction.2847|188|0更新时间：2024-05-07
- 摘要：Due to the physical structure of single-chip CCD or CMOS color image sensor, there is only a single color component at each pixel position. As a result, color interpolation or color demosaicing is required to reconstruct the other two color components. General linear interpolation method may blur the image edge, and introduce color artifacts near edges. Although state-of-art methods may increase image quality, the computational complexity limits their applications. This paper presents a novel color interpolation method. It effectively increases the quality of interpolated images in three ways. At first, it performs interpolation in color difference space rather then in normal color space. Secondly, it introduces Rational Functions (RF) based operator instead of linear operator. And at last, a post-processing step is employed further to enhance the image quality. Because interpolation in color difference space takes into account the correlations between each color components, and RF based operator is edge-adaptive, this is the reason why the image quality can be improved using this method. In addition, this method is simple and calculating efficient in contrast to some of the state-of-art methods. The experiments compare PSNR and PESNR as while as MSE with the other three methods at the end of this paper.关键词：Computer image processing;Image sensor;Color interpolation;color space;color filter array;rational function2715|206|0更新时间：2024-05-07
- 摘要：This paper discusses an algorithm of color image enhancement that is based on a neuro-dynamical model of the ON-OFF neurons in the human visual system. An appropriate form of this model for color image processing is found and the advantages and disadvantages of this model are also analyzed through the computational simulations. Extensive computations demonstrate that this model can achieve a very good degree of color constancy with the selection of a proper passive decay rate constant. At the same time, the trade-off between image enhancement and the fidelity of chromatic rendition is determined by the space surround constant. This neural system performs well on the enhancement for the natural scenes with complex contexts. However, because this neural system is a model of receptive fields of the ganglion cells in the human retina and still based on“gray-world”assumption, and it fails to handle the violations of the gray-world assumption. It shows that this model is still not comprehensive enough to describe the complex visual system and have some restrictions in practical application. Finally, we discussed the possible future improvement of this model.关键词：Computer image processing;image enhancement;ON-OFF neural networks3327|197|0更新时间：2024-05-07
- 摘要：The current approaches to feature point correspondence are based on the hypothesis that the displacements for feature points between consecutive frames are small. That is, there is very short time interval between successive two frames in image sequences. These approaches are difficult to find the corresponding feature points when the time interval is large. In the paper, a new approach to feature point correspondence is proposed. It includes two steps. First, approximate location of feature points of moving targets can be obtained using polar-exponential grid sampling and log-polar coordinate mapping. The parameters of feature point position, including translation, rotation and scale, are got in log-polar coordinate and Cartesian coordinate. Then the corresponding feature points can be got with available traditional means, because feature points corresponding when the time interval is large turns into feature points corresponding when the time interval is small by the first step of the approach. The best advantage of the method is simple and efficient. In the paper, the principle of the method is introduced and the result of experiment is given. It turns out that the problem of feature point correspondence can be finally solved when the time interval between consecutive frames in image sequences is large, and that the first step of the approach is enough to resolve feature point correspondence of plane object moving.关键词：Computer image processing;Image sequence;Polar-exponential grid technique;Polar-log coordinate transform;Feature point correspondence2796|193|0更新时间：2024-05-07
- 摘要：After the image coordinates of seed objects are computing and camera system composed of the single camera and a mirror are calibrated, searching corresponding points of seed centroids in the single-frame image is a key technique. In this paper, considering the distributing traits of the seeds in the single-frame image and the characteristics of genetic algorithm such as simple, current, robust and so on, a point-pattern matching method of the seed objects in the image based on the genetic algorithm is proposed. Firstly, by the fuzzy clustering method, the image points of seeds are distinguished to two groups, and then the genetic mechanism is established between the two point groups in the image, and the corresponding points are searched in the single-frame image. The experiments show this method can search corresponding points between two point groups of objects very fast and correctly, it is good for calculating the seed position in the space.2632|199|0更新时间：2024-05-07
- 摘要：This paper carried out a study on motion quantitative analysis, mainly aimed at two dynamic medical images (heart-echocardiograph and vessel-cineangiography), with the regularization method of computer vision and the clinic knowledge of cardiac wall and coronary vessel motions, and thus provided doctors with a useful tool to the diagnosis of cardiac disease. Our method first carries out a block-matching algorithm using the Lagrange operator to estimate the initial motion vectors. Then for the application of heart motion analysis, a Spring system model is employed to optimize the global smoothness; for the application of blood flow analysis, an assumption of local motion consistency is given, which offers a primal solution of this problem. An application software has also been completed. Experimental result shows that our algorithm is more efficient than the existing methods.2707|167|0更新时间：2024-05-07
- 摘要：Face contour extraction is important in facial feature extraction and in model-based coding. For the face boundary, classical edge detection techniques will fail to exploit the inherent continuity of face boundaries. As the shape of face boundary is not uniform and exhibit low overall curvature, geometric active contours are an attractive choice for the extraction of face boundaries, but the original geometric active contour model still has no way to characterize the global face shape. In this paper a new method that incorporates prior shape information into geometric active contours for face contour extraction is proposed. As in general a human face can be treated as an ellipse with a little shape variation, the prior face shape is represented as an elliptical curve. By combining the prior face shape with the geometric active model proposed by Chan and Vese, our improved geometric active contour model can capture face contour depending on both the image edges and the prior knowledge of face shape. Moreover, our model is implemented using variational level set approach, thus the transformation parameters (such as the rotation angle in plane) that maps the face boundary to the prior shape can be roughly estimated simultaneously. The experimental results show the efficiency and effectiveness of our method.关键词：Computer image processing;Geometric active contours;level set;Face contour extraction3783|186|0更新时间：2024-05-07
- 摘要：In order to reduce the spectral distortion existed in the fusion images generated by wavelet transform method, this paper developed a new fusion method called direct addition to process the low frequency information. The method directly added the original resized TM image that was not decomposed and the reconstruction images generated only with the high frequency parts of the decomposed SPOT image. The analysis of the spectral distortion and image entropy showed that the direct addition method was better than the general wavelet transform methods. Because the direct addition method does not need to decompose the multi-spectral images, its fusion speed was faster than the general wavelet transform methods. The spectral quality comparison of wavelet transform and IHS transform showed that wavelet transform is more suitable than IHS transform for the fusion of images with different spectral response ranges and for the fusion of images with little correlation. The spatial quality comparison showed that wavelet transform was excelled at the fusion of images contained medium scale or small scale texture characteristics.关键词：Computer image processing;Image fusion;wavelet transform;IHS transform2458|203|0更新时间：2024-05-07
- 摘要：Augmented reality is the merging of synthetic sensory information into a user's perception of a real environment. One of the most important tasks for every augmented reality application is to make the virtual scene and real scene merged correctly and become an organic whole. The implementation lies heavily on the cooperation between augmented reality hardware and software technology. Based on the analysis of traditional virtual-real registration technology which lies heavily on position tracking and camera calibration, we investigate the application of affine representations in augmented reality, and put forward a new method to get the relative depth of each rendered point. By studying the application of image analysis technology in AR technology, a numerical background expression method is put forward to facilitate the feature point extraction in the virtual-real registration process. We have performed both dynamic and static virtual-real registration experimentation using affine theory. The experiment results indicate that this method can effectively reduce virtual-real registration errors in applications of augmented reality.3986|202|0更新时间：2024-05-07
- 摘要：With data collected from the analysis of the characteristics of brush, water, ink and paper, this paper intends to standardize the painting process of Chinese ink and wash (sumi-e), making the process controllable and predicable. It also tries to classify the rice paper and the painting techniques and defines the perceptual artistic presentation of sumi-e in a digital term. On the basis of the above analysis, this paper intends to establish a three-dimensional paper fibre model to achieve computer simulation of the one-stroke effect in wash painting. Moreover, the influence of paper fibre is taken into consideration in order to obtain a more diversified and more authentic simulation. This accordingly provides a unique platform for real-time dynamic simulation of realistic ink diffusion in that the result ant diffusion images reflect the texture and the global change of gray tones as well as local temporal variations specific to a particular paper. At last, examples are presented demonstrating the capability of the proposed technique for handling different ink and paper properties.关键词：Computer graphics;Chinese ink wash drawing;Paper fibre model;texture;computer simulation3040|229|0更新时间：2024-05-07
- 摘要：The paper proposes a valid blind image-watermarking algorithm. Especially, the algorithm is also effective for some kind of images that have a lot of white pixels. These special images usually can produce a lot of overflows after being embedded the watermark. Furthermore, the magnitudes of these special images are bigger than usual images in some positions after being transformed. So, these images are hard to get the balance between robustness and transparentness. The method is based on Fourier (FFT) transform. The algorithm embeds the watermarking signal modulated on some fix positions that keep some distance from each other. Furthermore, the paper adds some decimals to the image-signed and choose some interrelated regions to extract the watermarking info during detecting, which reducing some data damage caused by saving the image-signed. Experimental results and attack analysis are given. For different images, the algorithm embeds different parameters during embedding watermarks in order to identify watermarks clearly. For the same image, the parameter of extracting is smaller than the one of extracting and the area of extracting is bigger than the one of embedding. The experiments show that it is transparent, robust, accurate and cost-efficient. Especially, the algorithm is strong to withstand the attacks of cutting, noising, compression, erasing and translation.2552|170|0更新时间：2024-05-07
- 摘要：Digital watermark is invisible mark embedded in digital image, video or audio documents, which can be detected to protect copyright of multimedia data. The thesis describes a watermarking scheme based on CL multi-wavelets. After the CL multi-wavelet transformation of an image, most energy of the image is mainly concentrated on its lowest resolution sub-image. Moreover, most energy of the lowest resolution sub-image is further concentrated on its first component. Because lowest resolution sub-image can survive from some transformation such as compression, this scheme can improve robust of embedded watermark. But the invisibility of watermark is reduced at that time because of the sensitivity of the lowest resolution sub-image to HVS(Human Visual Model). The position to embed watermark is selected with a trade-off between robustness and invisibility. And preprocessing of the watermark image is here for declining the difference between the watermarked and original image. To deal with attack of counterfeit, it uses not additive method but replacement algorithm when embedding watermark. The correlation between the original and extracted watermark is computed to tell whether there exists watermark in the image. The performance of the proposed watermarking is invisible to HVS and robust to a variety of signal distortions, such as JPEG compression, filtering, cropping and incorporating attacks.2525|214|0更新时间：2024-05-07
- 摘要：Real-time rendering of landscape has attracted growing interesting in recent years. The existent accelerating methods of 3D terrain rendering have made their contributions to terrain visualization, but, they have themselves drawback, especially for large scale terrain visualization at interactive rate. Large scale terrain visualization contains the following contents: terrain multi-resolution representation, terrain and texture paging, terrain and texture LOD control and efficient data store and retrieve. Based on the research about management and real-time rendering of large scale terrain data, one method of generating multi-resolution terrain model is improved, which combines the representation of terrain multi-resolution model with view-dependent control. A new method of storing and retrieving terrain data by quadtree is proposed. Furthermore, an all-in-one terrain visualization system which integrates adaptive triangulation, dynamic scene paging and data updating is realized. At last, some experiments have been presented to verify the work done in this paper.关键词：Digital elevation model(DEM);Level of detail(LOD);Multi-resolution model;View dependent2709|272|0更新时间：2024-05-07
- 摘要：The modeling of traditional realistic image generating is on the basis of polygon, which has some deficiency. For example, drawing precision depends on the degree of detail and the method of dissecting, quantity of dissecting and crossing computing should be needed during the generating of realistic image, the procedure of sculpt is complicated, and so on. A representing method based on discrete points is proposed in this paper. First we pick out some points on each curved surface in the scene. Then sort and store them. At the same time, a transparence-refracting algorithm is proposed and realized. The key of this approach is defining the checking line segment. Only the points in the sequences that are corresponding to the pixels on the checking line segment may be on the refracting line and the points in other sequences cannot be on the refracting line. Thus a huge of points that need not participate in computing are abandoned and computing quantity is reduced. The experiment results show that the representing method and the transparence-refracting algorithm can reflect the natural transparent phenomena virtually and accurately. At the same time, this new method can eliminate lines or faces easily and manipulate flexibly, it's data-structure and sculpture are simplified to makes the adding, deleting and modifying curve faces easily. In conclusion the method based on discrete points adapts to the complicated scene and cartoon.关键词：Computer graphics;Discrete points;Depth sorting;Checking line;Transparence;Refracting line2446|171|0更新时间：2024-05-07
- 摘要：Vehicle intelligent auxiliary drive system is the important part of the intelligent transport system. Using computer vision technique to solve this problem is a challenge work for every researcher in this field. The lane keeping and distance measurement based on the computer vision technique is the key problems of the auxiliary drive system. This paper describes the 2D reconstruction and distance measurement algorithms based on the monocular vision. The image is captured by CCD camera on the highway. The key thought is that establishes restriction functions for lane lines, tracks and reconstructs the lane line, gets the necessary parameters for lane keeping. Then based on the 2D reconstruction and geometric restriction, the distance between the vehicle and the obstacle can be measured with a single view. This paper also analyzes the parameters of the lane and gives the rule for lane keeping for automation navigation. All the algorithms are finished by VC++6.0. The experiments on highway in Sichuan province and Chongqing city show that the lane keeping algorithm is real-time and accurate enough to utility; and the accurate of monocular measurement algorithm can be accepted by the collision avoidance. The whole system can work well with 80km/h on the highway. To improve the real-time quality of monocular measurement algorithm is the next work.3705|228|0更新时间：2024-05-07
- 摘要：In this paper, an improved JPEG-LS algorithm is proposed to meet the needs of higher near-lossless compression ratio for remote sensing images. First, the threshold magnitude for run-length coding is increased and a prediction based on the control of local gradient is employed to extend the average run-length value, and this results in higher compression ratio. Second, because the subjective quality of the reconstructed image compressed by JPEG-LS at higher ratio degrades quickly, an error-modifying approach is added to the run-length coding area and a smoothing filter is applied to the reconstructed image. Thus the rebuilt image when observed by human eyes is much better and the objective quality measured byPSNRis higher as well. Finally, Golomb-Rice Coding method, which is a part of the original JPEG-LS, is adopted to encode residual errors. This can control the maximum absolute error of every pixel value in the rebuilt image and therefore keeps the great advantage of the original JPEG-LS in the sense of error control style. Integrating the above measures into a software package, under the same limit of ±7 for maximum absolute error of every pixel value, not only the compression ratio is somewhat increased, but also the human visual perception is obviously improved in the comparison with JPEG-LS. In our experiments, PSNRis about 1～3dB higher.3108|201|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