最新刊期
卷 14 ， 期 3 ， 2009
- 摘要：Human activity analysis is receiving increasing attention from computer vision researchers. This interest is motivated by a wide spectrum of applications, such as surveillance, man-machine interfaces, video coding, and so on. It has been a hot research in image analysis, psychology, and neurophysiology. This paper gives an overview of the various tasks involved in image analysis field. We focus on three major areas:（1) development history of the activity analysis,（2) important and novel ideas, （3) open problems for future research. Research about activity analysis was originated 30 years ago. In the recent years, increasing attention has been paid to this field. The two basic problems of activity analysis are the representation and the recognition of activity. This paper reviews the existing algorithms based on the two basic problems. The representation of the activity can be classified into two classes: the methods based on low-level image information and the methods based on the high level human model. And there are two kinds of activity recognition algorithms: template matching methods and state space methods. Finally, some research challenges and future directions are discussed.关键词：activity analysis;action representation;action recognition;state space method;template matching method8420|546|0更新时间：2024-05-07
- 摘要：A new gait recognition method based on information fusion of multiple kinds of features and views is proposed in this paper. Through the background subtraction and shadow elimination, human motion silhouettes are obtained and gait features are extracted using pseudo-Zernike moment, wavelet descriptor and Procrustes shape analysis. The gait recognition is accomplished through information fusion of multiple kinds of features and views on feature level and decision level. The method is evaluated on the CASIA gait database and received comparative high correct recognition rate. The experimental results show that our approach has efficient recognition performance.关键词：gait recognition;pseudo-Zernike moment;wavelet descriptor;Procrustes shape analysis;information fusion6785|348|0更新时间：2024-05-07
- 摘要：In this paper, a simple but efficient algorithm based on tensor subspace learning is proposed to reduce the dimensionality of high-dimensional silhouette data for human action recognition. For image sequences of each action, they are projected into a low dimensional subspace so that both spatial and temporal properties of the action are preserved. Further, a nearest-neighbor action recognition is carried out basing on Hausdorff distance. Two experiments for action recognition and robust test have been carried out to testify the effectiveness of introduced tensor subspace learning.关键词：human action recognition;tensor subspace learning;principal components analysis（PCA）;locality preserving projections（LPP)5873|327|0更新时间：2024-05-07
- 摘要：Identifying the minutiae correspondences between two fingerprints is crucial for minutiae-based fingerprint matching. With respect to the matter, this paper proposes a new algorithm for fast reference minutiae pair selection to alignment the minutiae set. Introducing the ridge information into the selection process and finding a novel net structure, we specify the most similar pairs based on each of these net structure pairs. Experiments on two sets of fingerprint images show that the proposed algorithm is fast with high accuracy.关键词：fingerprint recognition;minutiae;minutiae set alignment;reference minutiae pair5106|323|0更新时间：2024-05-07
- 摘要：This paper presents a factor that could direct to different fingerprint identification methods. With this factor, different fingerprint images can be classified easily and be suited for its best recognition method. Furthermore, this factor is based on four descriptors which are called: texture regularity descriptor, texture energy centralization descriptor, texture parallelism descriptor, and texture uniformity descriptor to analysis images from both local and global, frequency and time domain Finally, these four descriptors are joined together to classify the images by AND gate. Experimental result shows that because of the method proposed in the paper, it is not necessary to do complex preprocess before recognition and it can improve the accuracy and speed of fingerprint identification.关键词：classification factor;recognition direction;texture descriptor5598|345|0更新时间：2024-05-07
- 摘要：Using Gabor filter to enhance dry fingerprint image is an effective method. The key step is estimating the parameter of Gabor filter. In this paper, the direction and frequency of Gabor filter is estimated by Radon transform, and then fingerprint image is enhanced by the Gabor filter. Experiment results for some dry fingerprint images in FVC2000 fingerprint bases showed that the proposed method can produce effective enhancing with high processing rate.关键词：Fingerprint enhancement;RADON transform;Gabor filter;binarization5523|299|0更新时间：2024-05-07
- 摘要：Orientation field estimation is an essential preprocessing step and it is crucial for the efficiency of the fingerprint recognition algorithm. In this paper, we propose a Grid-Interpolation Model. It partitions the fingerprint plane around singular points with discrete grids, and builds the nonlinear relationship between orientation field and singular points. The orientation of grid points is adjusted by global information of the fingerprint, which is much different from the conventional method. Two experiments are carried out on FVC2002 and FVC2004 databases and the results indicate that the proposed model can reduce noise more efficiently and has more robust performance than the conventional method. Especially as for fingerprints with bad quality, it also can give satisfactory estimation.关键词：fingerprint recognition;orientation field;fingerprint analysis;singular points;fingerprint model5784|311|0更新时间：2024-05-07
- 摘要：Aiming at the real time image restoration, a novel image restoration algorithm using cellular neural networks is proposed in this paper. First, based on the classic regularized image restoration algorithm, the edge direction constraint is introduced with easy hardware implementation. Then, suitable template parameters of cellular neural networks are designed based on the CNN energy function. The cellular neural network is applied to implement the regularization function of the traditional model of degraded image. Simulation results show the efficiency of the new algorithm, its restoration results are better than the least square image restoration method with constraints and another image restoration algorithm base on CNN. Because of the easy hardware implementation and parallel realization, the real time image restoration can be realized with the new algorithm.关键词：cellular neural networks;image restoration4838|300|0更新时间：2024-05-07
- 摘要：In this paper, a modified bound constrained regularization model was proposed for image denoising and deblurring. In the model, we choose the variable exponent linear growth function proposed by Levine as the regularization term and choose the regularization parameters adaptively according to the image local feature. It not only preserves the advantage of the total variation regularization model, but also reduces staircase. Then, we introduce some bound constraints to the modified model, such as limit the range of u, to improve the quality of the restored image. At last, the constrained minimization problems are solved by primal-dual active-set method, essentially a semi-smooth Newton’s method. The numerical results show that our method is feasible and valid.关键词：image restoration;bound constraint;primal-dual active-set method;regularization method3988|276|0更新时间：2024-05-07
- 摘要：Bilateral filtering is an edge-preserving filtering technique; it employs both geometric closeness and radiometric similarity of neighboring pixels. A 2-dimensional implementation of the bilateral filter and its float closeness, however, are not suitable for hardware implementation and computationally expensive. In this paper, a modified bilateral filtering algorithm is proposed. It improves the bilateral filtering in three aspects: designing integer geometric closeness, restoring image in 1-dimensional and adaptively choosing control parameter of radiometric similarity. Experiments show that the modified bilateral filtering can achieve better performance.关键词：bilateral filter;Gaussian noise;edge-preserving;image restoration6375|352|0更新时间：2024-05-07
- 摘要：A hybrid Fourier-wavelet denoising method is proposed in this paper. The main steps of the proposed method are as follows. The noisy image is firstly denoised in Fourier domain. Secondly the remaining noise is removed in wavelet domain. The remaining noise is colored. In order to consider the correlation between wavelet coefficients of colored noise the Gaussian scale mixture model for images wavelet coefficients is used. Experimental results show that the proposed algorithm improves denoising performance efficiently.关键词：image denoising;wavelet transform;Fourier transform3689|279|0更新时间：2024-05-07
- 摘要：In the conventional video coding system, the distortion is always measured by Mean-square-error (MSE). However, the MSE-based distortion fails to measure the subjective difference of videos. As a result, the property and interest of human visual system (HVS) should be considered by video coder. This paper proposes a new perceptual coding method aiming at compressing the luma coefficients in terms of the eyes sensitivity. By using the pre-quantization strategy, it discards the imperceptible information and improves the compression performance of video codec without contaminating the video subjective quality. The experiment result proves that this perceptual coding method can efficiently reduce the output bit rate of AVS reference software by 8%~40% while guaranteeing the quality of decoded video.关键词：perceptual coding;human visual system;visual redundancy;Webers law4200|283|0更新时间：2024-05-07
- 摘要：By enriching redundancy of the contourlet transform, it is possible to weaken pseudo-Gibbs phenomena in the process of image de-noising by thresholding. In order to remove noise from image effectively and quickly, by enriching redundancy of the contourlet transform and avoiding too much data, a method for image de-noising based on non-subsampled pyramid contourlet transform is proposed. The method decomposes noisy image using nonsubsampled LP for multi-scale, and decomposes sub-image using critical sampled DFB, then performs scale related threshold for shrinkage, finally reconstructs de-noised image. Experiments compared with other related methods show that the proposed method, on the PSNR values of the de-noised images, yields improvements up to 1dB over original contourlet transform; on the time consumption, costs half less than other improved methods.关键词：image de-noising;Contourlet transform;nonsubsampled filter banks3772|296|0更新时间：2024-05-07
- 摘要：A novel support vector regression method—FRSVR(fuzzy robust support vector regression) based on traditional ε-SVR is proposed in this paper. First, a solution for support vector regression with arbitrary cost function is given. Second the properties which a cost function should have in order to construct a robust support vector regression are discussed. Then a family of cost functions is introduced. In the training procedure of FRSVR, outliers can also be detected in terms of different training error ranges between normal examples and outliers using fuzzy c-means algorithm (FCM). Through iteration, FRSVR is obtained. Since it is based on ε-SVR, various optimization methods for epsilon support vector regression can be used to solve FRSVR. In the experimental part of the paper, FRSVR is applied to process emulsified oil combustion flame images such that outliers therein can be detected and removed, then flame shapes are accordingly segmented. Experimental results show that FRSVR performs better than ε-SVR and ASBF filter in the sense of removing outliers and enhancing the generalization ability.关键词：outlier;support vector regression(SVR);fuzzy clustering4384|323|0更新时间：2024-05-07
- 摘要：Thresholding is one of the widely used and efficient techniques for image segmentation in digital image processing. Threshold selection is crucial to thresholding. The maximum between-cluster variance algorithm based on L2-Norm, which was proposed by Otsu, is one of the most famous methods. And the maximum between-cluster average deviation thresholding algorithm based on L1 -Norm is simpler and has good performance. The two-dimensional maximum between-cluster average deviation thresholding algorithm, which has high accuracy of segmentation and good resistance to noise, has better performance than the maximum between-cluster variance algorithm, but the two-dimensional algorithm requires a large amount of computation and is impractical in applications. In this paper, two fast recursive two-dimensional maximum between-cluster average deviation thresholding algorithms are proposed, whose computational complexities are only O(L2), while the computational complexity of the original1 algorithm is O(L4) . Using those two recursive algorithms, the results and processing time of the two-dimensional maximum between-cluster average deviation thresholding algorithm are given, which are compared with the original algorithm. Experimental results show that both of those two recursive algorithms can greatly reduce the processing time, which is only 0.1% of that of the original algorithm. Currently the proposed algorithms have been used in automatic infrared target,vehicle license plate and fingerprint recognition system.关键词：image segmentation;threshold selection;two-dimensional maximum between-cluster average deviation;recursive algorithm4560|328|0更新时间：2024-05-07
- 摘要：The small target detection in infrared image is one of the crucial techniques for the precision guided weapon. Using wavelet and contourlet multi-scale analysis to detect small targets in infrared image has certain limitations. non-subsampled Contourlet transform (NSCT) is a new extension to contourlet transform. It is flexible multistage and shift invariant multidirectional. This paper presents a new small infrared target detection approach, which is based on nonsubsampled contourlet transform. This method picks up the texture characteristic from the maps after nonsubsampled contourlet transform, and calculates the distance between the characteristic vector and the center characteristic vector. This can gain a correlative multi-scale distance map. The histogram of distance map is employed for the automatic selection of threshold value. The experimental results are shown, and compared with the results based on wavelet transform in terms of detection precision and speed . The results show that the method of small infrared target detection based on nonsubsampled contourlet transform can precisely detect the small infrared target and it is better than the method based on wavelet transform.关键词：small infrared target detection;non-subsampled contourlet transform;texture analysis4684|301|0更新时间：2024-05-07
- 摘要：In order to acquire accurate passenger flow information by locating and tracking moving pedestrians accurately in image sequence, a novel approach of head feature extraction based on target region matching is presented. Deferent from the common methods based on dense disparity image to obtain the disparity of the head region, the method in this paper is based on the idea of “segmentation before matching”, i.e. the reference image is segmented firstly by monocular image processing and the candidate head regions will be acquired. Then these candidate head regions will be taken as the target regions to be directly used in the correspondence regions searching and matching to obtain the head disparity. Finally the disparity of the candidate head regions is used to extract the depth and perspective feature of the candidate head regions to remove the false head regions. The performance test and experiment results show that the method proposed in this paper has the advantage of higher precision of disparity extraction and better real-time performance as well as the great importance to extract the pedestrian heads feature with high degree of distinction to effectively eliminate the false head regions so that the accuracy of the passenger flow detection can reach over 90%.关键词：stereo;image processing;block matching;target region matching4410|301|0更新时间：2024-05-07
- 摘要：Object tracking using single feature often leads to a poor robustness. In this paper, a new object tracking algorithm based on multiple features fusion is presented. to alleviate the affection of object deformation and partial occlusion, it analyzes and describes the color, texture, edge and motion feature of the object using a consistent histogram model, in order to conquer the distractions in the complex background, these features are rationally fused in the framework of Auxiliary Particle filter to obtain a more satisfying approximation of the posterior distribution of the object states. A new method to estimate the fusion coefficient is also proposed to improve the fusion result. Experiment results show that our algorithm can efficiently cope with both rigid and non-rigid objects, outperforms single feature based object tracking algorithms, and has a high robustness in complex background. The comparisons with other multi-cue tracking algorithm also show the validity of the proposed algorithm.关键词：object tracking;multi-cue tracking;probability fusion;Auxiliary particle filter3988|285|0更新时间：2024-05-07
- 摘要：Horizontal line can be used to precisely describe the shape of a certain region. Line segment table is the data structure of line. In this paper, a fast algorithm of line segment coding and seed filling based on line segment table is proposed. Every element of the table describes a horizontal line. The five parameters of the table are L, xL, xR, y and F. They denote the symbol of relationship of up and down rows, the x coordinates of left and right end, the y coordinate of the row and a flag for line type respectively. Base on line segment table, the connectivity of rows and columns can be calculated conveniently. Fast algorithm of line segment encoding follows the steps of line searching, labeling and sorting. The fast algorithm achieved by the concept of line segment table is 1/2 faster than contour filling. Results show that the data structure of line segment table is much more effective and flexible for data storage and data processing.关键词：line segment table;line segment encoding;seed filling;pixel labeling;contour filling;fast algorithm4379|269|0更新时间：2024-05-07
- 摘要：To visualize 3D images, people usually attempt to extract certain isosurfaces from 3D images to represent approximately boundary surfaces of some structures within these 3D images. Considering that some boundary surfaces could be well approximated by certain isosurfaces, but many boundary surfaces within 3D images actually cannot be well approximated by any isosurface, it becomes an important problem to distinguish which boundary surface in 3D images could be well approximated by certain isosurfaces. In this paper, we address this issue and provide a practical and effective method solve the problem.关键词：boundary surface;isosurface;zero-crossing points;3D edge-detection;clustering3683|254|0更新时间：2024-05-07
- 摘要：With the advantage of Contourlet and the disadvantage of single feature fusion rules, a multi-polarization SAR image fusion algorithm, based on product of local energy and regional standard deviation, is proposed. First, an adaptive threshold method is adopted to preprocess the images. Then, multi-polarization SAR image fusion is implemented in Contourlet domain. According to the characters of Contourlet coefficients, local energy fusion rules and weighted averaging method is adopted for low-frequency sub-band fusion rules. The Product of regional energy and local standard deviation as a fusion rule is adopted for high-frequency sub-bands. The result of real multi-polarization SAR images fusion show that this algorithm can provide satisfactory fusion performance.关键词：SAR image fusion;Contourlet;speckle reduction;adaptive threshold;feature product-based fusion rules3667|260|0更新时间：2024-05-07
- 摘要：A new algorithm was presented to fuse color visual image and infrared(IR) image aiming at detecting weapons hidden underneath a persons clothing. This algorithm combined dual-tree complex wavelet transform (DT-CWT) with RGB, HSV and LAB color space. DT-CWT was used due to its better properties than other wavelet transform in fusion. These properties include shift invariance, directional selectivity, limited redundancy, perfect reconstruction and computational efficiency. The fused image maintained the high resolution of the visual image, incorporated any concealed weapons detected by the IR sensor, and kept the natural color of the visual image. The feasibility of the proposed fusion technique is demonstrated by some experimental results.关键词：color image;IR image;image fusion;DT-CWT;concealed weapon detection3678|203|0更新时间：2024-05-07
- 摘要：Firstly, the Harris corner detector is used to detect corners, a method to describe the direction character of the corners is given and the keypoint descriptor of SIFT(scale invariant feature transform) is introduced in this method. Secondly, Euclidean distance is performed to get exact matching in the corner set. Finally, an easy and effective method to avoid wrong matches is discussed. This method is effective on brightness invariant, movement invariant and rotation invariant in the matching of digital city image sequences. This method is proved to be effective by realistic experiments, and of high practicability.关键词：digital city;image matching;Harris corner detector;point feature;scale invariant feature transform(SIFT)4414|335|0更新时间：2024-05-07
- 摘要：A method of affective annotation of images based on multi-feature is proposed in this paper. First, the corresponding relations between visual features (color, texture and shape) and emotion in an image are discussed, and the relations results are provided by orthogonal experiment, research and reference documents. Then, proper algorithms have been selected and modified for abstracting the features of color, texture and shape, by way of fuzzy inputs. Finally, an affective space of emotion quantification and an approach of fuzzy approximately reasoning theory is proposed to annotate affective. The experimental results are given to show that the proposed approach achieves a higher degree of precision, by asking 20 persons to estimate 100 images affective in contrast with the system. At the same time, it is approved that this proposed affective quantified space is provided with scientificity and practicability. It can be applied at decoration, electronics teaching, picture indexing and emotion calculation.关键词：affective quantification;feature abstraction;affective annotation;mapping3244|274|0更新时间：2024-05-07
- 摘要：One of the most challenging work in non-photorealistic rendering research field is to studies of the simulation of Chinese ink wash drawing. By analyzing the Chinese ink wash drawing material and ink-wash diffusion mechanism, a new method based on genetic algorithm is presented to model the diffusion effect. The method first simulates the diffusion effect by binary gene encoding, initializing population and gene mutation. It constructs a list of mapping that describes the correspondence of Minkowski distance to the rate of gene mutation so that gene mutation can happen suitably by itself. As a result, modeling the diffusion of Chinese ink wash drawing can be finished automatically. All kinds of ink wash diffusion effects can be realized by some parameters, such as the scale of population, paper cell fiber threshold and so on; Furthermore, to improve the diffusion effect made by the simulation system above, the model uses Ashikhmin algorithm and Gaussian Distribution to simulate the characteristic of xuan-paper to make the diffusion effect with more fidelity. Experimental results demonstrate that this method simulates the typical effects of Chinese ink wash drawing successfully.关键词：non-photorealistic rendering;diffusion effect;genetic algorithm;xuan-paper;texture synthesis3777|254|0更新时间：2024-05-07
- 摘要：Lines and bodies are two important spatial objects in 3D space. Research on topological relations between lines and bodies have great value in many application fields such as mining, military and transportation. In order to analyze the topological relations between 3D lines and bodies especially complicated lines and bodies correctly and efficiently, a group of sample spatial objects are defined. Then nine-intersection model is used to describe the topological relations between spatial objects. Using nine negative conditions between lines and bodies, eleven types of topological relations between sample lines and bodies are deduced. Then the composed reasoning method is used to deduce the topological relations between compound lines and bodies and there are altogether nineteen types. Finally the algorithm to analyze the topological relations between complicated lines and bodies is deduced. After detailed analysis and discussion, the completeness of topological relations between lines and bodies is proved. Therefore, the algorithm can guaranteed the correctness and greatly simplify the analysis of the topological relations between complicated lines and bodies.关键词：nine-intersection model;topological relations;topological analysis;completeness of topological relations;lines;bodies3835|284|0更新时间：2024-05-07
- 摘要：Morphing of point-sampled geometry is one important research area in the field of computer animations. Based on spherical parameterization, we put forward a robust morphing of point-sampled geometry. Source and target models represented by point-sampled geometry are first parameterized onto a sphere, respectively. After aligning the corresponding features of two models on their spheres, two spheres are projected onto a common rectangle-parameter domain and the correspondence between sample points on the two models is built using this rectangle domain. In order to preserve the geometric details of point set surfaces, the absolute geometry of the in-between models is computed by means of Laplacian operator and is dynamically up-sampled using a moving least square method so as to eliminate the cracks. Experiment results demonstrate that our algorithm can preserve the geometric details very well and produce a smooth transition sequence.关键词：point-sampled geometry;spherical parameterization;Laplacian operator;morphing;moving least square surface3830|246|0更新时间：2024-05-07
- 摘要：The paper presents the framework and key technique of a MAR system in allusion to the restriction in working range and operability of general MAR system. Then the author emphasize the framework and advantages in mobile application of Java3D which is an image rendering technique used in MAR system. Finally the feasibility of the MAR system in the paper is validated with experiments. Moreover, the wide area application perspective and its potential commercial value are also expected.关键词：augmented reality;mobile angement reality(MAR);Java3D4758|297|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