最新刊期
卷 7 ， 期 4 ， 2002
- 摘要：This paper proposes an approach to feature point correspondence of image sequence based on neural networks. We formulate the correspondence problem as a constrained optimization problem and propose a 2D Hopfield neural network to solve it. The design of energy function of neural network has ranged over the results of visual tracking and the condition of occlusion. Therefore, it can solve the problem of error correspondence resulting from current existing methods. The correct correspondence of the first three frames is very important for the point tracking. This paper develops a 3D Hopfield network to handle the correspondence of the first three frames and proposes a cost function of motion smoothness to formulate the energy function of 3D Hopfield network. Experiment on a real image sequence demonstrates the feasibility of the approach.2486|343|0更新时间：2024-05-07
- 摘要：The corner is an important local feature of image. To avoid the disadvantages of using the single feature to detect the corner points, a new algorithm based on multi feature is proposed in this paper. In this algorithm, intensity feature and edge feature are contributed to corner detection. First, a fast adaptive SUSAN principle, which utilizes the local gray level feature directly, is proposed for detecting the candidate corners. This improved method can detect features, such as corners, edges and intersections, in different contrast image automatically. For detecting the corners on blurry edges, the candidate corners would include some edge points as a result of reducing the detection threshold. These candidate corners, which include true corners, some edge points and a few false points, are arrayed along the boundary trend by the method of edge element. Through these arrayed points, the angles between approximate straight edge lines are calculated to be as the criterion of determining a corner. Those edge points are removed since they have not significant discontinuous changes in the direction of boundary, i.e. the angles of them are not acute enough, and the false corners due to quantization also are removed by our method. After these steps, the true corners are reserved. The experimental results showed this corner detection method having good capabilities of detection and localization in different contrast image.3312|409|0更新时间：2024-05-07
- 摘要：In this paper, we present a simple but effective editorial system for facial customization and expression animation. Provided with frontal and side facial images and a general 3D polygonal facial mesh with muscule vectors, facial feature lines are fitted by snake automatically. variation interpolation are performed on the general model to customize 3D virtual face model. Then, we generate non crack facial texture mosaic with multiresolution spline. And the so generated virtual particular face looks highly realistically. According to its parameterization representation of imbedded muscular vector, the virtual face can make different expressions after we edit relevant parameters. Owning to its topological invariability, we can realize 3D morph between different particular faces based on triangular image metamorphosis. The system can be realized on PC plat easily and quickly and it is very practical.3129|420|0更新时间：2024-05-07
- 摘要：In this paper, the monthly digital NDVI image with 8×8km spatial resolution is used to study the correlation between NDVI change and climate factors, such as temperature and precipitation, and the rules for the regional distribution of NDVI change form 1983 to 1992 in China. It shows that the driving factor of land cover change in Northeast China, East of Inner Mongolia inland and Tibet plateau is precipitation, and the leading driving factor of most South China region(southeast monsoon), Huanghuai plain and West Xinjiang region is temperature. From southeast to northwest, it shows difference driving factors and influencing intensities and it can be found obvious strap regularity. Based on these results, it is pointed out that there are three image forms of NDVI change driven by climate in China(driven by temperature, driven by precipitation and driven by temperature-precipitation), and it indicates further the regional difference of NDVI change in China.4117|641|0更新时间：2024-05-07
- 摘要：Generally, the analyzed results from remote sensing data are uncertain and multi solution, which is determined by the characteristics of global surface information being multi dimensional and infinite. Therefore, remote sensing information has some degree of definite statistical characteristic, but as well as holds the high randomness and complexity, which generally behaves as mixture density distribution in feature space. In allusion to the complexity of statistical distribution of remote sensing information, in this study we firstly introduce into the finite mixture model and its expectation maximization(EM) algorithm for decomposing the mixture distribution into finite parametric density distributions in order to simulate or approach the whole mixture distribution. By the model it should be firstly assumed that whole distribution could be separated into infinite parametric density distributions, then by EM iterative computation the maximum likelihood parameters of each proportional distribution can be estimated. Furthermore, the finite mixture model and its EM algorithm are extended to clustering algorithm for remotely sensed data. By the experimental case, the EM clustering algorithm is synthetically compared with conventional statistical clustering algorithm. The results show that the EM algorithm has several particular advantages such as self adaptive decision for clustering number, extensibility of prior knowledge integration and free initialization, etc.3593|452|0更新时间：2024-05-07
- 摘要：This paper introduces the theory and method for remote sensing images fusion. Based on the multi band wavelet theory, a scheme of remote sensing images fusion based on image feature is made in this paper. In this algorithm the number of bands for wavelet transformis decided on the resolution ratio among remote sensing images, high resolution image to be fused can receive the low frequency ingredient by the multi band wavelet transform. By displacing the low frequency ingredient and the low resolution images with each other according to some relations, the new low frequency ingredient of high resolution image is formed. Then the fusion image through the inverse multi band wavelet transform can be obtained. this scheme makes best use of the information on remote sensing images to be fused and prevents the loss of image information. This method is experimented by using material images. In the proposed method all the spectral characteristics have been preserved, and the definition and the spatial quality of image have been improved. Experimental results including the fusion of SPOT panchromatic image and SPOT multi spectral images, SPOT panchromatic image and Landsat(TM) images are presented. Compared with other methods, the proposed scheme is much better and it possesses more adaptability.3443|439|0更新时间：2024-05-07
- 摘要：Positron emission tomography(PET) is a noisy functional medical imaging model. In this paper a fully Bayesian PET reconstruction method is presented for combining a segmented anatomical membrane a priori. The segmented anatomical membrane a priori is based on the fact that the radiopharmaceutical activity is similar throughout each region and the anatomical information can be obtained from other imaging modalities such as CT or MRI. The prior distributions are formed as some kind of Markov random field. Due the non convex and the hyper parameters in the prior, it is difficult to use point estimator such as maximum a posteriori(MAP). So we used Dynamic Markov chain Monte Carlo posterior simulation method to get a minimum mean square error(MMSE) estimator which update the hyper parameters as well as density data. The variances and credit area of the reconstruction results can be easy gotten by MMSE. We compared the reconstruction result of ML, MAP and MMSE, and find that the segmented anatomical membrane a priori exhibit improved the noise and resolution properties and Dynamic Markov chain Monte Carlo is mostly suitable for fully Bayesian reconstruction.2637|343|0更新时间：2024-05-07
- 摘要：Image inpainting is an important research topic in the area of image restoration, its objective is to restore the lost information according to around image information,which can be used to restore old photo. Because edge information of image is very important ,it is prior to take edge information restoring into account while designing image inpainting algorithm. Based on total variational model, a digital image inpainting algorithm is presented in this paper. Total variational model can simulate low level human vision, during the process of image inpainting, which can restore edge features in image. Numerical experiment results show that this model can restore the information in inpainting area fairly well, but it will be disturbed by the size of inpainting region, so, an information propagation method is used first to shrink the size of inpainting region.2810|408|0更新时间：2024-05-07
- 摘要：A new space-adaptive regularization method for blind image restoration, which combines the idea of priori blur identification is presented. This new technique first identifies the point spread function(PSF) by using alternating minimization iterative algorithm. Then it restores the image based on the identified PSF using anisotropic regularization. The main difficulty in blind image restoration is insufficient information, which demands full utilization of priori knowledge of image itself and imaging system. This algorithm utilizes the piecewise smoothness of both the image and the PSF, and it simultaneously makes use of the concept of anisotropic diffusion, which carries out space adaptive regularization according to the orientations of the image and the PSF. This new method's efficiency is demonstrated by numerically blurred images. It can get better restoration images than space adaptive regularization(SAR) method, and converge faster than space adaptive anisotropic regularization(SAAR) method2474|396|0更新时间：2024-05-07
- 摘要：In this paper, we introduce an algorithm of global motion estimation based on multi line features which are different characters and which can be distributed regularly in the image region. Global motion estimation attempts to find the shift of cameras such as translation, zoom and rotation. Multi line segments are obtained by wavelet edge extraction algorithm and Hough transformation. Straight lines used in estimating motion parameter are selected under the control of selection rulers. Global motion parameter is obtained by comparing straight lines between the original referent image and the distributed images. Its application for electronics image stabilization is also presented. The distribution of an image sequence can be move away by filtering the trace of frame motion obtained by global motion estimation. The image sequence could be obtained by compensated to image frames. It is satisfing in experiment that global motion parameter can be calculated correctly and robustly depend on several straight lines. Un stabilization can be shift by averaging motion vector estimated by multi-line segments.2568|376|0更新时间：2024-05-07
- 摘要：Lifting scheme, the new hotspot of wavelets research, can be used to implement the existent wavelets and to construct completely new wavelets. Seen from the theory of lifting scheme, any transform constructed by lifting scheme is some kind of generalized wavelet transform. What's more, many digital signals are from the integral average of analog signals over certain interval or area. Considering the above two facts, the concept of average interpolating is presented in this paper. The average interpolating wavelet of 2 degree is constructed and is applied to image coding. After the wavelet transform for image data, Optimal Bit Allocation, which combines embedded quantization and entropy coding of binary system arithmetic coding together, is applied to get an optimal coding result by reaching the least distortion with the expected bits number. Finally, an image coding result is got as good as, if not better than, that of the D9/7 recommended by JPEG2000 draft. The lifting steps of the average interpolating wavelet is much simpler than that of the commonly used D9/7, which means the computational resources consumed here is much less.2590|436|0更新时间：2024-05-07
- 摘要：There is a wide foreground for 3D modeling concerned with faces under Internet environment. Panoramic range image and its corresponding panoramic color image are used to describe 3D faces in this paper. Based on it, 3D morphing of faces under the internet environment is introduced, and is attempted with reasonable tradeoff methods by using virtual reality modeling language in world wide web browser such as Microsoft internet explorer with cosmo player plug in. The results show that: with reasonable designs such as proper automatic mesh subdivision and realistic texture mapping, the describing data could be made as small as possible to fit for applications in the internet; meanwhile realistic efforts could also be guaranteed. Such a research could supply technical supports for virtual communication, virtual broadcasting and tele education in the Internet. It is expected to improve the interface and to increase the interesting of the Internet by using such a technology.2286|349|0更新时间：2024-05-07
- 摘要：clipping is one of basic problems in computer graphics.Some condtions relating to clipping will be met when people draw grid contour,such as some sheet or ringy region without date point, region's boundary,some region with inaccurate interpolation point. However, at present,some famous clipping algorithms home and aboard are almost impossible to process grid contour's clipping after it is smoothed because time complexity is too high and region is too complicated.In this paper, an algorithm suit for the clipping of grid contour to solve the problem is presented. The algorithm may be summed up the following three steps:①filling the clipping region with -2.0 by calculating the point of intersection between contour and clipping region and sorting the point of intersection and partnering the point of intersection.②recording the direction of trace of contour from every boundary of clipping region.③increaseing the end condition of contour that begin from some boundary of clipping region and to own boundary or other clipping region' s boundary.As a result,the algorithm passed the test of practical large data.The examples shown in the paper and other applied examples demonstrated that the method of clipping is fast and effective for complex region of grid contour.2763|444|0更新时间：2024-05-07
- 摘要：As a system for collecting, storing, analyzing, processing and managing geography spatial data, GIS's spatial data are mainly composed of vector data and grid data. The existing GIS systems are mostly based on vector data and have a great ability to establish and manage vector database. But they have poor ability to process grid data, especially to organize, schedule, store and manage mass grid data, without mentioning the ability to manage and integrate the multi source, multi scale and multi phase image data. This paper does a deep research and a thorough discussion of mass image database and its realization and application. The paper mainly discusses the research background, system organization and implementation of mass image database(IMIDB) based on relational database system(RDBMS) and component object model(COM). It focuses on the logical design(including the selection of Database, the organization and the storage of remote sensing data); Web publishing architecture of remote sensing images in IMIDB and the application of its Component Object Model. Finally, a test of this system is done and the result of the test is given. From the result, we give a conclusion that the system has reached the aim of the design.2542|346|0更新时间：2024-05-07
- 摘要：Temporal GIS and spatio temporal data model are currently research focus in geographical information science(GeoScience). Land cadastral information system is one of the most typical application systems in TGIS. In this paper the spatio temporal process of the land cadastral information system is analyzed. A kind of section fast indexing(SFI) in spatio temporal data model of base state with Amendments(BSA) and a method of storage factors of variable granularity(SFVG) are presented. By importing SFI and SFVG, the inquiring efficiency of system is markedly improved. Then the spatio temporal concept, data structure and the method of implementation for Land cadastral information system are discussed. How to organize data, build topological relations and storage data in database are emphatically discussed. At last, a software package of dynamic land cadastral information system based on above model and methods is introduced. The other important technology used in dynamic land cadastral information system is self developed Spatial Database Engine(SDE). The main functions of the system are simply introduced. The dynamic land cadastral information system is practically used for more than two years in many provinces, which shows that its design is advanced, efficiency high and operation stable.2856|338|0更新时间：2024-05-07
- 摘要：In this paper, an algorithm to generate digital terrain model rapidly is presented. The algorithm is based on the constrained delaunay triangulation(CDT) algorithm. The algorithm achieves a high performance by efficient managing data for the digital terrain model by quad tree, and by reducing the calculation of the algorithm by quad tree. At first, the discrete points of the digital terrain model are distributed to different leaf nodes of the quad tree; then, the points in the leaf node of quad tree are triangulated by delaunay triangulation algorithm; at Last, meshes in the nodes which are neighborhood in space are coalesced together to generate a new smooth mesh. The algorithm is high performance because in the coalition which only need to deal with the points in the hull of the mesh in each node of quad tree. Moreover, the constrained edges and constrained polygons can be integrated to the mesh quickly by using quad tree structure. In the paper, the test results of the algorithm and the analysis to the test results are also presented. Moreover, a test picture of the algorithm is illustrated. In the last part of the paper, the time cost analysis of the algorithm and the spatial characteristics analysis of the algorithm are presented. The test results show that the algorithm not only has excellent performance but also is robust. The time cost analysis of the algorithm shows that the expected time of the algorithm is O(n log( n )), and the analysis of the spatial characteristics of the algorithm shows that algorithm can be adaptive to different spatial distributed situation of the points set to be triangulated. Meanwhile, the constrained edges and the constrained polygons can be integrated into the triangulated mesh efficiently with the help of quad tree structure of the algorithm.2641|408|0更新时间：2024-05-07
- 摘要：A scheme about the technology for the seamless composition between the presenter image shot by vidicon and the virtual environment image coming from computer in virtual studio has been investigated in this paper. A algorithm has been presented in which the plane wireframe mesh about the video object would be constructed by closed B spline based on the editable points (EP) coming from unbiased partitioning sample points on the segmentation edge about the video object. This algorithm has been used for the construction of the vision realistic graphics for the presenter in the virtual studio. The experiment has been accomplished to validate the feasibility of the algorithm submitted in this paper, which is carried through capturing on image sequences of single presenter by shooting from vidicon. The experimental result has shown the validity of the algorithm for the simple 3D reconstruction of the video object by fitting the segmentation edge curve and proved that the algorithm is effective and self adapting in the building of the graphical mesh.2703|445|0更新时间：2024-05-07
- 摘要：Augmented Reality(AR) is a growing area in virtual reality research. It merges the virtual object or other additional information generated by the computer into the user's view of the real world. It is helpful to apply AR technique to the scene planning for architectural designers or beauty spot developers. In this paper, a typical AR application system--Scene Planning is introduced to demonstrate this idea.To combine the virtual buildings and real scene, the geometric coherence and illumination coherence problem must be sloved. The local 3D reconstruction and interactive camera calibration are used to resolve the geometric problem. For the illumination problem, the interactive and untomatic method is combined,and the shadow of the virtual object is also taken into account. The human's interaction is also involved during the above process. By evaluating the result image synthesized from various view angles, the designer can estimate and optimize the layout of virtual scene intuitively, and eliminate the inaccuracy during the planning process. Finally, some work under improvement and future research are also discussed.3227|397|0更新时间：2024-05-07
- 摘要：The object based rate control problem arises along with the video object concept proposed by MPEG 4. The video objects can be coded in synchronous or asynchronous way, but latter one has more advantages in low bit rate coding and it is also proved by the experiments of this paper. This paper proposes a new rate control algorithm which can achieve a constant bit rate when coding asynchronous video objects. Firstly, the algorithm estimates the total target bits at one time assuming that the video plane from the same video objects of the same coding type and close in time are coded with a similar number of bits. Then the algorithm to allocate the target bits to the coding video objects at the time in order to achieve constant distortion ratios between pairs of video object planes adopts an allocation policy called constant distortion ratios. This paper also proposes a post processing technology to solve the problems of composition picture quality brought by asynchronous video objects. The experimental results show that the algorithm can efficiently control the coding bits of asynchronous video objects and achieve high subject and object quality.2504|395|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