最新刊期
卷 15 ， 期 3 ， 2010
- 摘要：PM（perona-malik） is a classical nonlinear image diffusion model with adaptive properties of image smoothing and edge preserving. It is transformed to geometric surface smoothing, edge preserving variational model using variational level set method based on the analogue of normal vector of surface and general image intensity in this paper. The energy functional includes three parts. The first part is a data term using the square of the difference between the Heaviside function of initial level set function and the one of the evolved level set function. The second part is a regularizing term based on geometric curvature and the third part is a penalty term enforcing the level set function to satisfy the property of signed distance functions. Numerical examples validate its abilities of surface smoothing and edge preserving.关键词：surface diffusion;PM model;surface denoising;edges preservation.4856|396|0更新时间：2024-05-07
- 摘要：H.264 KTA increases the coding performance at the cost of high computational complexity.Its adoption of adaptive interpolation filter（AIF）and 1/8 pixel MV resolution causes that fractional pixel motion estimation（FPME）exerts more significant influence on the computational complexity of video coding.In this paper，we propose an algorithm to speedup FPME by using uni-model characteristics and the MV correlation of fractional pixel motion compensation.According to experiments，our algorithm achieves much higher speed than other fast FPME algorithms with neglectable R-D lost compared with F帧/s algorithm.3974|312|0更新时间：2024-05-07
- 摘要：In order to take effective copyright protection for 2D engineering drawings，the paper presents a reversible watermark embedding algorithm based on difference expansion for 2D engineering drawings. The algorithm works on the feature of poor relevant of vertex coordinates of lines in engineering drawing, embedding watermark information by modifying the vertex coordinates of lines. Because of minute revision, there isnt any visual redundancy. During information extracting, the algorithm can restore the original engineering drawing accurately, and meets the requirements of high-precision machining, and also realizes the blind extraction. The experimental results show that the algorithm achieves good results, and easy to implement, it can be applied to the copyright protection of engineering drawings, content certification,information secret transmission and other fields.关键词：2D engineering drawings;difference expansion;reversible watermark3122|285|0更新时间：2024-05-07
- 摘要：In order to improve connexity and accuracy in skeletonization, a new skeletonization method using template and neighborhood information is proposed. Firstly, we smooth the bilevel image and calculate the neighborhood sum of every point in vein line to differentiate fringe points and medial axis points, then find some medial axis points through the template matching and obtain some discontinuous axle wires by deleting all isolated points；finally, we obtain the vein skeleton by using algorithm of maximal neighborhood point tracking from the ending points of the axle wires. Experimental results show that vein skeleton extracted by the algorithm is smooth and stable, and superposes on actual axle wires. At the same time, the algorithm is provided with scale and rotation invariance，and strong anti-noise ability, so it is effective to extract vein skeleton.关键词：vein recognition;skeleton extraction;the maximal neighborhood point;template matching;medial axis point4284|377|0更新时间：2024-05-07
- 摘要：数学规划 鞍点规划 颗粒图像 鞍点关键词：mathematical programming;saddle point programming;granule image;saddle-point3557|298|0更新时间：2024-05-07
- 摘要：In the light of the detection of multiple vehicles in the traffic video sequence the paper presents an edge detection based random walk algorithm for accurately detecting and segmenting the multiple vehicles in the video. First, the regions of the moving vehicles are detected by the background differences and the edge information, and the skeleton structures are extracted from the detected vehicle regions. Secondly, the effective pixel markers, which are obtained from the skeleton structures, are taken as the object seeds in the random walk algorithm. Finally, the automatic detection of the vehicles is realized by using seeds to segment multiple vehicles in a random walk way.关键词：random walk;background difference;Canny;skeleton;vehicle detection4286|432|0更新时间：2024-05-07
- 摘要：Based on the distance ratio function to measure the gray level of a single pixel and the similarity between two pixels, this paper presents a novel image edge detection algorithm based on the measuring of medium truth scale. Two thresholds and restraint on non-maxim value in domain are applied to the algorithm. Simulation results show that the algorithm can successfully remove noise, preserve the image edge details and is more effective and has wider applications compared with the classic algorithms.关键词：image edge detection;measuring of medium truth scale;similarity degree3616|607|0更新时间：2024-05-07
- 摘要：The detection of linear feature is one of the most fundamental tasks in computer vision and image processing. Principal lines are the most visually prominent lines in the images that people mostly care for. This paper presents a novel algorithm for automatic detection of the principal lines in digital images. The feature edge pixels are first detected from the image with an image segmentation technique. Then they are approximated with short line segments, which are further clustered using some similarity metric. Each cluster of segments is fitted with a straight line in an iterative manner. The principal lines are determined according to the validity measurement. We demonstrate two practical applications including automatic image recovering and automatic photo composition to show the applicability of the principal lines in image processing.关键词：principal line;line detection;segment;clustering;fitting;validity4761|357|0更新时间：2024-05-07
- 摘要：Shadow detection is a key problem in object detection, object tracking and visual surveillance. In this paper, a new shadow detection algorithm is proposed based on fuzzy Markov random fields. The shadow detection problem is regarded as a problem of searching the optimal labeling of the total foreground pixels. The background abstract algorithm is used to find out the shadow and foreground pixels in the image. The fuzzy Markov random fields are created after the calculation of the shadow probabilities, the foreground probabilities and the membership functions. Bayesian principle, maximum a posteriori(MAP) estimation, iterated conditional mode(ICM) algorithm are used to search the optimal fuzzy Markov random field. The result of the shadow detection is obtained by defuzzifying the fuzzy Markov random field according to the maximum membership principle. Experimental results demonstrate the performance of the proposed algorithm.关键词：shadow detection;fuzzy Markov random fields;Bayesian principle;MAP(maximum a posteriori) estimation;ICM(iterated conditional mode) algorithm3547|364|0更新时间：2024-05-07
- 摘要：Conventional rotational invariants of Tchebichef moments are represented by rotational invariants of geometric moments, therefore the drawbacks of geometric moments, such as high degree of information redundancy and sensitive to noise, are inevitable. New rotational invariants of Tchebichef moments are proposed in this paper. Uusing the properties of falling factor，the proposed moments are transformed to linear combinations of Tchebichef central moments which can be evaluated by Tchebichef moments directly. Experiments show that our descriptor outperformed the conventional descriptor in rotational invariants.关键词：discrete orthogonal moments;Tchebichef polynomials;rotational invariants3421|312|0更新时间：2024-05-07
- 摘要：In this paper, a face recognition method, which utilizes an information fusion for color images and supervised neighbor preserving embedding, is presented for improving the perform ance of face recognition. First, Gabor transformation is used to extract information per channel of color image respectively, and then canonical correlation analysis is utilized to fuse extracted Gabor features. Supervised neighbor preserving embedding is used to reduce dimensionality. Finally, nearest neighbor classifier is used to classify reduced features. Experiments are carried on XM2VTS and FRAV2D color face databases, and utilize principal component analysis, linear discriminant analysis and supervised neighbor preserving embedding to reduce dimensionality of Gabor features on gray method and multi-channel feature fusion method. These results show that the combination of multi-channel information fusion and supervised neighbor preserving embedding can improve the performance of recognition system.关键词：gabor feature;canonical correlation analysis;supervised neighbor preserving embedding;information fusion4171|387|0更新时间：2024-05-07
- 摘要：A binocular Helmholtz stereo algorithm based on iterative dynamic programming method is proposed here and applied for measuring objects with highlights. In this algorithm, we first rectify the Helmholtz reciprocal image pairs. Then the end points of matching segments are detected. Lastly an iterative dynamic programming matching method is proposed to search for the matching points and reconstruct the depth of surface. Ray-tracing binocular Helmholtz images with highlights are used in our experiment and the disparity results show its good performance for reconstructing objects with highlights.关键词：Helmholtz;stereo matching;highlight;dynamic programming3306|307|0更新时间：2024-05-07
- 摘要：For the CCD serial star images, while the space target is moving, the background is still moving because the CCD flat is moving. The global movement of background stars and the local movement of the space target both exist in the star-sky images, which add difficulty to dim target detection. In order to detect dim space target with high efficiency, a star-sky image registration algorithm is put forward. First is the pretreatment, including radiation asymmetry rectification and stray-light rectification. Then, for each frame of image, centroids of the brightest stars are computed. The stars which can radiate in every frame are used as control points to compute the global motion parameters. Finally, the star-sky image is transformed into a form of star centroid coordinate matrix. And the background stars can be eliminated by using the global motion parameters. The algorithm omits image resampling and transformation steps, and thus saves abundant of computing time. The registration image can eliminate the background stars and the dim target can be easily detected from the registration images.关键词：star-sky image;image registration;feature points extraction;feature points match3334|388|0更新时间：2024-05-07
- 摘要：Depth-image-based rendering (DIBR) is a key technology to the advanced video applications. In order to improve the image quality generated by view warping, this paper proposes an arbitrary view generation method based on DIBR and image inpainting. The depth image is processed by morphological operations to reduce the holes generated during image warping and to smooth the object contour in the target image. The view warping equation is utilized to generate an arbitrary view. The image inpainting method is employed to restore holes inside the warped view. Depth cue is incorporated into the cost function for texture estimation. The pixels in the most match block corresponds to the holes in the target image are copied to the holes. During the restoration process, luma-first strategy is applied to adapt to various color sampling format. Experimental results show the proposed algorithm has superiority in both subjective details and PSNR statistics.关键词：depth image based rendering;view warping;image inpainting5325|373|0更新时间：2024-05-07
- 摘要：To solve the computational demands of spectral clustering approach when applied to SAR image segmentation，a multiscale method is proposed.It consists of three steps，i.e.coarsening，initial segmentation and refining.First，a sequence of smaller graphs，each with fewer vertices，is constructed from the SAR image.Second，spectral clustering is applied on the smallest graph to obtain initial segmentation.Third，the initial segmentation is refined scale by scale to get the final segmentation of the SAR image based on the posterior probability of classification which is estimated by the mixture model.Finally，experimental results demonstrate the effectiveness of the proposed method.关键词：spectral clustering;mixture model;posterior probability of classification;multiscale segmentation;SAR image3747|348|0更新时间：2024-05-07
- 摘要：The key to target-based SAR image compression is to corporate ATD algorithm with image coding. A SAR image encoder embedded with a multiresolution CFAR （constant false alarm ratio） detection algorithm in wavelet domain used db4 is proposed in this paper. Target areas are detected before quantization. Significant wavelet coefficients mask is derived based on db4. The sequences of target areas are encoded with a higher bit rate than those of background. Compression and decompression are done on MSTAR target chips； the quality parameters for target areas are achieved； comparison is made with a conventional SAR image coding algorithm of W/TCQ. Experiments show that SNR of target areas using TIC algorithm with low bit rate are higher than that of using W/TCQ algorithm and context information is preserved.关键词：synthetic aperture radar;wavelet;target detection;image compression3591|370|0更新时间：2024-05-07
- 摘要：Monopulse is a technique with which information concerning the angular location of a target is obtained by comparison of signals received in multi-channels.It is widely used in tracking radars.This paper proposes a valuable monopulse imaging algorithm for airborne radar forward-looking imaging.As is prored by experimental results, compared to the real-beam imaging method，this algorithm can provide remarkably improved images，which are difficult to be obtained by SAR or DBS.关键词：monopulse;forward-looking imaging;sum-difference beams;angle estimation5004|360|0更新时间：2024-05-07
- 摘要：Electronic image stabilization（EIS）technique has been widely utilized for removing the unwanted motion fluctuations from video sequence.It is one of the primary development directions of the future image stabilization technique.This paper firstly introduces the basic principles of EIS and its architecture.After giving an overview of the state-of-the-art of key techniques including motion estimation，motion correction and image compensation in EIS，we list a number of general means to assess the performance of video stabilization process.Finally，the status about the application of EIS is summarized and its development trend is analyzed.Survey results show that EIS technique possesses promising application prospect and great market demand，as research on EIS technique has turn into engineering application.关键词：electronic image stabilization(EIS);motion estimation;motion correction;image compensation;performance assessment8557|596|0更新时间：2024-05-07
- 摘要：The segmentation of abdominal CT series is a challenging task due to problems such as blur edges, large variance among individuals and small sample sizes. In this paper, a hybrid 3D surface segmentation algorithm based on a multi-resolution integrated model and missing data recovery technique is proposed. The appearance models to characterize the texture features around surface points are established, and the"confidence level （CFL）"for each point is defined. For the points which have high confidence, segmentation is accomplished by active image searching and model deformation. While for the points which have low confidence, instead of using unreliable edge information, data recovery technique is applied based on a statistical deformable model and available high confidence points. The experimental results demonstrate that the Hybrid-MISTO achieves the lowest segmentation error compared with a variety of state-of-the-art techniques such as Snake, ASM, and MISTO.关键词：abdominal image;image segmentation;multi-resolution;active shape model;data recovery5224|356|0更新时间：2024-05-07
- 摘要：Directed against uneven illumination of image，this paper proposes a threshold image segmentation algorithm based on edge detection.According to this method，the continuity of spatial distribution of the objects，background and the illumination has been considered to determine the threshold by using the information of the edges gradient.Then，the average of the maximum and minimum value near the edge detected by edge detection algorithm is used as the local threshold for the pixels on the edge.Using surface fitting，the threshold surface is constructed by the edge-threshold control points for image segmentation.The experimental result shows that the algorithm is efficient，and its Edge-Border Coincidence is higher than other typical algorithm.The ability of Anti-fuzzy of this algorithm is stronger and it has overcome the weakness of the edge-based method which is easy impacted by noise.关键词：edge detection;uneven illumination;image segmentation;surface fitting5347|381|0更新时间：2024-05-07
- 摘要：Traditional methods for fast patch matching can deal with only image patches with predefined shapes. However, in some cases （e.g. image completion） the patch shape is irregular and different from patch to patch due to destroyed or missing data. In this paper we propose an efficient method for accurate irregular patch matching. We formulate irregular patch matching as a problem to find the minimum weighted SSD （WSSD）, and the shape of patches is controlled indirectly with the weights of pixels. In this way all patches can be taken to be rectangular, and then the computation of WSSD can be accelerated with fast Fourier transform （FFT）. By taking advantage of the property that large area of patch should be padded with zero. We improve the FFT algorithm and further improve its performance without sacrificing accuracy. At the end of this paper we take image completion as an example to show that WSSD is a more general measure of patch similarity than SSD, and can serve as a uniform framework for various applications that involve image patch matching.关键词：irregular;patch matching;image completion5142|393|0更新时间：2024-05-07
- 摘要：This paper presents an image-based human face fusion algorithm，which can be used in medicinal beauty and digital entertainment areas.This paper combines facial features extraction with image fusion algorithm，so that we can automatically obtain synthesis results without using 3D facial models.Our system allows users to adjust the interpolation weights between source image and target image to control the similarity degree.Furthermore，we use Laplacian as constraints to reduce shadows near the nose.The experiment results show that our algorithm is efficient enough for practical demands.关键词：poisson equation;active shape model;facial fusion;image blending5309|347|0更新时间：2024-05-07
- 摘要：We present a hybrid model for 3D cartoon water modeling and rendering. First, we constructs some cartoon water forms and arranges them on 2D working region and map them onto 3D surfaces. The region covered with an obstacle object is detected on the 3D surface and then mapped reversely onto the 2D working region. The spray form is added if the flowing water forms collide with the mapped region. Our method is able to preserve the cartoon style of water forms during the interactions between the obstacle object and the cartoon water. Moreover, user can adjust the water forms, positions and the movement rate with ease. The real-time rendering is achieved by taking advantage of the graphics processing unit.关键词：cartoon animation;water;modeling;GPU;computer animation4956|356|0更新时间：2024-05-07
- 摘要：It is of much significance for disaster rescuing, evaluating, and propaganda, to do the research of the typhoon wind-field rules and realistic simulation. Based on physical characteristics of typhoon, this paper first builds the 3D asymmetry typhoon wind-field model, and through simplifying and dispersing, calculates different wind speeds in different conditions. Then combined with the real-time satellite images, the profile, intensity, track and moving velocity of typhoon are simulated. Further, considering the movement under different conditions for different rigid and flexible objects, and using these realistic rendering technologies, We simulated the dynamic typhoon scene successfully based on its characteristics finally. The simulation results are very realistic.关键词：typhoon wind-field;realistic modeling;quick rendering;physical characters5062|322|0更新时间：2024-05-07
- 摘要：This paper presents a new method to simulate large areas of near-shore waves. Firstly, the correct wave surface movement at appointed area in a period of time is obtained by solving Boussinesq type equations, which can describe the compositive phenomena of near-shore wave： refraction, diffraction, reflection and wave-wave interaction. Secondly, the movement of wave in finite time can be changed to infinite time by reconstructing the sequence of the waves height fields. Finally, in real-time rendering process the corresponding wave height fields are sampled by the view-dependent sea surface grid as vertex textures. The experimental results show that this method is realistic and effective. It is appropriate for real-time simulation applications.关键词：near-shore wave simulation;wave modeling;wave rendering;ocean wave simulation5615|332|0更新时间：2024-05-07
- 摘要：Its a growing concern to many researchers for real-time simulation of music fountains since they are fascinating decorations in cities and scenic spots which combines modern technology, music, and waterscape, etc. In this paper, a new approach for real-time simulation of large scale music fountain is proposed which incorporates fluid dynamic, particle system, music synchronization, and general purpose GPU computation. The fountain particles are dynamically synchronized with music data, and simple dynamic equations of particles are used to simulate realistic fountain behavior. The real-time update of particle attributes such as position, speed and the real-time rendering of particles are achieved through the availability of high parallel processing power of modern GPU. A new particle rendering method based on adaptive sampling in image space is also proposed for particle realism. The proposed algorithm has been verified and can be easily integrated into interactive applications for real-time fountain simulation.关键词：music;fountain;particle system;GPU;real-time rendering4725|376|0更新时间：2024-05-07
- 摘要：Molecular interaction field （MIF） is an important tool used in the research of molecular structure and function and has become a conventional method in drug design and molecular docking. Yet it is difficult to visualize and analyze molecular interactions with existing graphical representations of proteins such as VDW, molecular surface and CPK due to three dimensional nature of protein interaction distribution. A novel method for volumetrically visualizing and analyzing the MIF of a protein of known structure is proposed. Beginning from a proteins molecular structure, we compute an MIF after performing a molecular dynamics simulation on the protein and then visualize the MIF volume. A two-pass method for designing transfer functions is proposed specifically for producing informative images that reveal the MIF volume features. The volume rendering results are used to analyze HIV-1 protease with biologically meaningful result being derived.关键词：volume rendering;molecular interaction field;electrostatics;Van der Waals;drug design;molecular docking;protein4698|332|0更新时间：2024-05-07
- 摘要：Structure reconstructed from motion is usually distorted due to accumulated error and calibration error from extrinsic parameters of video cameras, although the re-projected error sometimes can be small. This paper proposes a novel approach to structure from motion with the geometry conditions existing in the scene, such as the trajectory of a video camera lying on a straight line. With this new method, distortion of the reconstructed results can be removed. We also propose a method to automatically select the optimal coefficients between the original cost and the constraint cost, through which the best reconstruction is obtained. The implemented examples demonstrate very precise structure and motion recovery, which prove the effectiveness and robustness of the proposed method.关键词：structure from motion;path constraint;multi-view3862|368|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