最新刊期
卷 15 ， 期 10 ， 2010
- 摘要：Phase retrieval refers to the use of direct measurement of intensity distribution to restore the phase and then reconstruct the wave function. A phase retrieval based on total variation with transport-of-intensity equation is presented to obtain phase information as much as possible. First the phase retrieval model is established in the case of uniform illumination, and then the problem is translated into minimum of energy functional with total variation regularization, and the finite difference Newton’s law is used to seek an optimal solution. Only the measurements of the spatial intensity of the optical wave in closely spaced planes perpendicular to the direction of propagation are needed and the sensitivity and resolution issues because of the requirements of interferometry techniques put on the spatial and temporal coherence of the optical source are avoided. The experiments show that information in the edge is kept when conducting phase retrieval.关键词：phase retrieval;transport-of-intensity equation;total variation;regularization3446|260|0更新时间：2024-05-07
- 摘要：Omni-directional camera has attracted more attentions in computer vision community due to its quick wide-angle imaging. But the great distortion makes the post-processing complex，which limits the algorithm applications of classical perspective camera.This paper deduces an unwarping algorithm from arbitrary ring omni-directional image to cylinder panorama based on the Taylor model of omni-directional imaging system. And we also present an unwarping method to produce a novel kind of panorama which is composed of N perspective planer images. Both two reprojection unwarpings do not depend on a particular model or accurate parameters of omni-directional system. In the unwarped panoramas，each pixel is assigned with a light ray in the 3D space，which makes the applications of stereo rectification and 3D reconstruction simple. Finally，the experimental results with the simulations and real data illustrate the validity of the proposed method.关键词：omni-directional image unwarping;cylinder panorama;N-profiles panorama4144|210|0更新时间：2024-05-07
- 摘要：The processing order of error blocks has an important effect on the improvement of concealed image quality, an improved adaptively enhanced temporal error concealment algorithm is thus proposed. The algorithm continuously estimates the order value of error macroblock(MB) using the received or concealed adjacent information, and only the error MB with the largest order value is restored at each time. When conceals an error MB, firstly we determine its partition mode and the concealed order of its sub-macroblock(SMB), then chooses the best motion vector(MV) for the error MB or SMB from the candidate MVs, by adaptively use the single-sided or double-sided overlapping boundary matching algorithm based on texture(TBSS-OBMA or TBDS-OBMA). The experiment results show that the proposed algorithm can improve the quality subjectively or objectively aspect in the same error environment, and the improvement of 1.16dB in PSNR can be achieved comparing to JM12.2.关键词：temporal error concealment;concealed order;boundary matching algorithm3088|198|0更新时间：2024-05-07
- 摘要：In image smoothing with diffusion equation, general methods are to construct difference equation of diffusion, then solve it with initialization and edge condition. These methods have low precision and diffuse error quickly. So a wavelet interpolation method is structured in this paper and is applied to solve the diffusion equation. We get a method to Alvarez model by two-dimensional wavelet interpolation method. Wavelet function possess better partial character, compared with finite difference method, wavelet method has higher precision, slower speed of error diffusion, and not is sensitive to time interval. The experiment shows the advantages of this method compared with difference method.关键词：diffusion equation;image smoothing;wavelet interpolation2902|230|0更新时间：2024-05-07
- 摘要：The classic variational image diffusion models for planar image processing are extended to image diffusion on implicit surfaces using intrinsic image gradients defined based on implicit surfaces and expressed by zero level set functions, which can preserve edges effectively during image diffusion. But the models using only intrinsic gradients of image intensity show staircase effects beyond edges. In order to reduce the staircase effects, a general hybrid variational diffusion model is presented including regularizers based on intrinsic gradients and intrinsic divergence. The TV model and PM model are implemented as examples to validate the formulation in reducing staircase effects.关键词：variational level set;implicit surface;image denoising;staircase effect;intrinsic divergence2942|202|0更新时间：2024-05-07
- 摘要：The object of this study is to establish a universal color image quality assessment method based on human vision function characteristics for perceiving noise in an image in order to assess color image quality including noisy color image. A comprehensive image quality assessment model called as the best quality assessment function for color images is established by means of four quality parameters of image quality，information entropy, averaging contrast, averaging gray and the standard deviation in a key region of an image. The maximum of the function corresponds to the color image with the best quality. The results assessed by this method agree with human vision perception. The method, which founds on auto-computation for four parameters characterizing image quality, is not only in no need any reference image but also can be used to assess the reference image quality. The method can be also used to acquire the color image with the best quality by means of a reference image through a gray transformation our called Zadeh-X transformation if the reference image quality is not good enough.关键词：noise perception;color image quality assessment;information entropy;averaging contrast;averaging gray3173|261|0更新时间：2024-05-07
- 摘要：Two kinds of compression-type vertex chain code are proposed. According to the characteristic that massive continuous two-vertex numbers are usually contained in an image boundary’s vertex chain code, in actual coding process, 1 and 3 are replaced by 0 and 9 separately, and 1 to 8 are used to express 2, a new dynamic vertex chain code is presented. The edge’s vertexe numbers of a figure are divided into four types: 1, 2, 3 and the combination of 1 and 3, based on which the top bit area and the low bit area of one byte are used to express the number and its continuous quantity of the code separately, a new compression type chain code is studied. Compared with the current vertex chain codes based on an efficiency-evaluation method, the results indicate that the new vertex chain codes not only have 50%～70% coding compression-ratio, moreover their coding or decoding efficiency is enhanced by 10%～40%, namely, their decoding real-time is better.关键词：vertex chain code;dynamic vertex chain code;compression-type chain code;comprehensive efficiency3061|168|0更新时间：2024-05-07
- 摘要：The LOF (local outlier factor) algorithm is a very distinguished local outlier detecting method,which assigns each object an outlier-degree value. In this paper,we present the two improvements of this algorithm. First, the two step improvements was introduced and their time complexity was analysed. Second,when the algorithm identified local outliers, it can consider spatial attributes and non-spatial attribute. The experiments have tested the executing time of the LOF algorithm and its improvements,the performance of computing synthetic and real data set. The experimental results show that is its improvements outperform the LOF algorithm in efficiency and performance.关键词：data mining;spatial outliers;reachability distance;local outlier factor (LOF)3495|240|0更新时间：2024-05-07
- 摘要：In view of characteristics for coke optical texture in micrograph, a segmentation algorithm，combining mean shift and edge confidence, is proposed. Firstly, the edge confidence of image pixels is calculated. With the edge confidence, the weighting function of mean shift algorithm is computed. The sampling points of feature space are weighted in order to improve the accuracy of detected modes. Secondly, coke optical texture is segmented preliminarily by iterating the weighted mean shift vector. Because that the number of clusters in initial segmentation is larger than that of the actual clusters, which may result in over-segmentation, combining conditions are set by the spatial distance and the average value of the edge confidence, which are used to combine regions of homogenous texture. The coke optical texture is finally segmented with the new combining conditions. Experimental results show that with the proposed algorithm the segmentation among different optical textures of coke is reasonable and effective, which offers a reliable foundation for the recognition of coke optical texture.关键词：coke optical texture;micrograph;mean shift;edge confidence;weighting function;image recognition3180|223|0更新时间：2024-05-07
- 摘要：The equivalence of the anisotropic diffusion and HAAR wavelet shrinkage in 2-dimension in a given condition has been proved. The anisotropic wavelet shrinkage based on the equivalence combines the merits of wavelet shrinkage and anisotropic diffusion. In this paper, applying the anisotropic wavelet shrinkage to multiresolution image segmentation is researched, one improved method to multi-resolution anisotropic diffusion image segmentation method is proposed, which is a multi-resolution anisotropic wavelet shrinkage image segmentation method. The anisotropic wavelet shrinkage is used to diffuse the pixels in the image, and the gray level of the neighbor pixels in homogenous fields tend to become the same value along with the scale’s increase while keeping the edge as far as possible. The scale space stack based on scale is built, and the segmentation to object is accomplished. It is an unsupervised method. Comparison experiments show that the method can locate the edge of the object exactly while processing the internal inconsistency of fields effectively, and the homogenous fields are merged perfectly to implement segmentation. The convergence speed of the method is faster than the multi-resolution anisotropic diffusion image segmentation method.关键词：image segmentation;scale space;anisotropic wavelet shrinkage2763|265|0更新时间：2024-05-07
- 摘要：A self-organization mapping based region Gaussian model algorithm is proposed, which can detect moving objects in complex background image sequences of the static scene. This algorithm combined single Gaussian model with region Gaussian model to form a cascading detection method, and the “competition & cooperation” mechanism was introduced in the step of region Gaussian model updating. This method can not only detect moving objects efficiently at the scene of shaking leaves, waving water surface and other complex background, but also guarantee the integrity of detected objects. Through the experiments of PETS2002 and Water Surface image sequences, the characteristic of proposed algorithm was validated in various environments with complex background.关键词：moving objects detection;region gaussian model;self-organization mapping;cascading detection2827|277|0更新时间：2024-05-07
- 摘要：The performance of mean shift algorithm using kernel histograms as tracking cues is always affected by illumination, visual angle and camera parameters. Kernel co-occurrence matrices (KCM) constructed on the concept of gray level co-occurrence matrix (GLCM) were used to represent the target model and the target candidate. Then those matrices were employed as the tracking features in mean shift tracking framework. Some improvements were made in the implementation of the algorithm. First, pixels on the opposite position of the current pixel were treated discriminately to depict the asymmetric characteristics of the object. Second, the KCMs of the target model and the target candidate were normalized to a same integer to improve calculation accuracy. Third, the computation of each pixel weight was modified to improve operation speed. The tracking results of several real world sequences with dark or changing illumination and partial occlusion show that the proposed algorithm can track the target effectively.关键词：mean shift;kernel co-occurrence matrix;visual tracking;weight computation2772|214|0更新时间：2024-05-07
- 摘要：An improved particle filter algorithm based on Camshift is proposed in order to track the moving target quickly and accarately. Firstly, under the particle filter framework, the algorithm improves the traditional target model and presents a novel target model, which fuses color and motion cues, to enhance the robustness and accuracy of target tracking. Meanwhile, in order to increase the tracking efficiency, an improved Camshift algorithm is embedded into the particle filter to rearrange the random particles, in which the particles moved toward the maximal posterior probability density of the target state. Experimental results show that compared with the traditional particle filter algorithm or Camshift algorithm, the proposed method can successfully cope with the situations of fast moving target or strong disturbances in the background, and achieve fast and robust tracking of the target.关键词：real-time target tracking;particle filter;Camshift;multi-cue fusion3224|277|0更新时间：2024-05-07
- 摘要：In order to enable TV tracking system to resist external interference when operating automatically, a second order accuracy scheme is proposed to automatically extract and fast track the moving targets. First, a double frame-difference method is presented to obtain the initial template of the input video image combined with the double gate approach of the quartered images. Then, we track the object effectively according to a trajectory prediction operator and adaptive feature matching technique. In a TV tracking system, the algorithm proposed in this paper has been applied in experiments of low-altitude flying targets tracking, which have achieved good tracking results. The results of the experiments show that the proposed method not only meets the real-time target tracking, but also can effectively improve the accuracy of the target identification.关键词：double gate;quartered image;frame comparison;trajectory prediction;feature matching3193|224|0更新时间：2024-05-07
- 摘要：To address the lack of explicit concepts and effective methods with learning in most hierarchical visual computational models, we propose a novel hierarchical generic object recognition sketch using neural sparse coding. Firstly, the learning strategies are embedded in a hierarchical system for generating several prototypes to model the characteristic of complex cell receptive fields. Secondly, based on a hierarchical sparse coding process, we present a hierarchical feature extraction method for generic object recognition. Finally, a simplified classifier is designed to achieve our goals in complex scenes according to the extracted robust features. Experiments on Caltech-101 demonstrate the effectiveness in our method and the more preferable results comparing with Serre’s show greater consistency in biological vision.关键词：biological vision system;hierarchical computational model;generic object recognition;neural sparse coding3506|200|0更新时间：2024-05-07
- 摘要：Feature spectrum descriptors give the rich, unified, workable expressions of the original characteristics of an image. In this paper, to solve the unilateral and interference problems in the application of one single feature spectrum descriptor, adaptive feature spectrum descriptor is proposed. Suitable features are choosen according to the characteristics of the image region itself. In order to test the performance of the feature spectrum descriptor, it was appllied to the content-based image retrieval. Experiments show that the new feature spectrum descriptor has good retrieval efficiency, and small time complexity.关键词：feature spectrum descriptor;texture spectrum descriptor;HSV histogram;content-based image retrieval2806|214|0更新时间：2024-05-07
-
Multi-view stereo reconstruction of volumetric human face via graph-cuts and dynamic patch structure
摘要：In this paper, a novel approach is proposed to reconstruct human face model from multi-view images with volumetric scene representation. This method focuses on improving the reconstruction results without the limit of the representative ability of 3D facial model samples. Our main contribution is to combine feature points based reconstruction and volumetric based reconstruction into the framework of Graph Cuts when applying facial reconstruction. Two considerations contribute to the improvement of final results. First, a dynamic patch based photo consistency estimation is designed to get the value of the photo-consistency constrain. Second, a dynamic weighted graph is constructed in order to avoid half voxel size error. We evaluate the performance on a real dataset, in which multi-view face images (8, 16, about 30) are captured by moving digital camera under changed illumination. The results indicate our approach have significant improvements.关键词：multi-view stereo reconstruction;graph-cuts;3D facial reconstruction;dynamic patch-based photo consistency estimation;dynamic graph structure3145|349|0更新时间：2024-05-07 - 摘要：A Curvelet based image fusion method is provided employing fundamental ultrasound contrast image with distinct tissue boundary information and corresponding harmonic ultrasound contrast image with distinct tissue lesion information acquired at the same frequency. Curvelet transform is applied to fundamental ultrasound contrast image and corresponding harmonic one at first in this method, then different fusion strategies are used for low frequency coefficient and high frequency coefficient to obtain ultrasound image with distinct tissue boundary and distinct inner details and to locate tissue lesion and tissue boundary. The experimental results show that this fusion method is promising.关键词：ultrasound imaging;contrast hanmonic imaging;image fusion;curvelet transform3698|190|0更新时间：2024-05-07
- 摘要：For SAR image waters edge detection, the traditional algorithm can not suppress speckle noise，so there are many false edges in the results. Based on Shearlet transform, a kind of multi-scale geometric transformations, which represent curves accurately and effectively, we propose an improved Shearlet transform. Combining improved Shearlet transform with clustering and snake model, a new method of waters edge detection for SAR image is performed. Experiments show that new method can not only reduce influence of speckle noise, but improve the integrity and accuracy of edge detection, and it is efficient and effective for SAR image waters edge detection.关键词：shearlet transform;SAR image;water;edge detection5414|222|0更新时间：2024-05-07
- 摘要：Due to the multifomity of the clutter statistic property in UWB SAR image, there are much false lines, discontinuity of road edge and big computation load when using traditional method based on one clutter statistical property, like ROA, to detect road edge from UWB SAR image. In this article, a novel method based on morphology is proposed which not only uses multidirectional and multiscale structures, but also utilizes original gray image information. First, the original image is transformed to binary image. Second, circular close-open operation is processed with the structures mentioned above to get consequential images. Finally with the original gray image information, the consequential images are fused to obtain the final results. Experiments show that the presented method, compared with the traditional edge detecting methods, can get fewer false lines, more accurate localization, better continuity and smaller computational load.关键词：UWB SAR images;edge detection;morphology;multi-structure3026|228|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