最新刊期
卷 19 ， 期 6 ， 2014
- 摘要：Iris image quality assessment is very important in iris recognition systems. Its motivation is to improve the performance of iris recognition by assessing the quality of iris samples. It is closely related to the modules of iris image acquisition, human-computer interaction, performance evaluation, adaptive iris recognition and so on. With the development of non-ideal iris recognition techniques, iris image quality-assessment methods have been widely studied in recent years. In order to give the researchers a comprehensive knowledge about iris image quality-assessment, state-of-the-art methods are reviewed. The trend, technical solution and at last, the potential pointers of the iris image quality-assessment methods and the quality driven iris recognition techniques are discussed. With the advent of the era of big data, large scale iris recognition will become the hot topic. The study of iris image quality assessment is of great significance to promote the processing capacity of the iris recognition system.关键词：low quality image iris recognition;iris image quality assessment;quality metric estimation;quality metrics fusion;spectral analysis4053|328|0更新时间：2024-05-07
- 摘要：With the rapid development of digital acquisition technology, media document are easily acquired and become an indispensable part of peoples' modern life. The powerful video editing software has made the video copy-paste forgery become more and more easy. Therefore, the appraisal of video authenticity has great significance. Direct extending the traditional image forgery detection algorithms to forgery video detection is computational expensive and time-consuming, moreover, the spatiotemporal consistency could not be preserved. In this paper, an intraframe copy-paste forgery video blind detection approach based on dense scale invariant feature transform(SIFT) flow is proposed. The proposed algorithm divides videos into sub-clips at the minimal content variation, extracts the keyframe as proxy frame. It detects the initial forgery region by matching SIFT keypoints, refines the forgery region by exploiting the SIFT key points dependence with mean shift region, and warps the keyframe detected region to the remaindering frames. The experiments showedthat the presented approach achieves one order of magnitude improvement in efficiency, and improves the mean detection accuracy by 20%. The proposed video forgery detection algorithm could efficiently detect the copy-paste forged regions within the video.关键词：intraframe copy-paste forgery;keyframe;dense scale invariant feature transform (SIFT) flow;video splicing4387|436|0更新时间：2024-05-07
- 摘要：To propose an image inpainting algorithm that can enable the computer to repair images with a larger damaged area and more complex structure information. Images with a larger damaged area and more complex structure information are repaired by imitating manual repair methods, and it has two steps: the division of damaged area and the repair of each block. 1)In the division process of the damaged area the matching degree of every two damaged boundaries are calculated first and the boundaries having the greatest matching degree are made into matching pairs.Then, the well-matched boundaries are directly connected, dividing the damaged region into different blocks. 2)In the repair process of each block, the transmission equation and diffusion equation of the BSCB algorithm are used to repair each block with the highest priority first. Second, our algorithm judges whether there is a block with second priority to be repaired, if yes, the boundary deletion algorithm is used to delete the redundant boundary lines, and the same method is used to repair each block; if not, the repair process is completed. Based on the above image inpainting steps, an image inpainting algorithm based on partition block of damaged region is proposed. The experimental results show that the proposed method can increase the PSNR value about 1.49 db and achieve better visual effect. The proposed algorithm is more suitable for repairing damaged images with a larger damaged area and more complex structure information compared with other three image inpainting algorithms.关键词：image inpainting;manual inpaiting;boundary lines;BSCB algorithm;structure information4539|456|0更新时间：2024-05-07
- 摘要：Since it is difficult to estimate the correlation noise distribution, a non-parametric estimation method for correlation noise is proposed. Non-parametric estimation method for correlation noise is proposed a according to the feature of correlation noise, a kernel density estimation-uniform distribution model based on optimal bandwidth is proposed. The impact of the noise model using parameter estimation method and non-parametric estimation method on the system performance in the transform domain Wyner-Ziv system is compared. The experiments results show that non-parametric estimation method can simulate correlation noise distribution and the established noise model using non-parametric estimation method can save rate about 10% at most at high bit rate for WZ frame. Non-parametric estimation method is an effective noise estimation method for TDWZ system.关键词：distributed video coding;transform domain Wyner-Ziv;correlation noise distribution;kernel density estimation;noise model3158|243|0更新时间：2024-05-07
- 摘要：By combining fractional calculus and duality theory, a novel fractional-order primal-dual model, which is equivalent with the fractional ROF model, is proposed. We theoretically analyze its structural similarity with the saddle-point optimization model. So the algorithms for solving the saddle-point problem can be used for solving the model. The primal-dual algorithm based on resolvent for solving the saddle-point problem is used for solving the proposed model. The adaptive variable step size iterative optimization strategy is used, which can improve the optimizing efficiency, and remedy the step size limitation of the traditional numerical algorithms. In order to guarantee the convergence of the algorithm, the range of the parameter is given. The experiment results show that the proposed fractional-order primal-dual model is effective in avoiding the staircase effect and preserving texture and detail information, and the adoptive numerical algorithm has faster convergence speed. In this paper, we propose a fractional-order primal-dual denoising model, which can be solved by a primal-dual algorithm based on resolvent. The experiment results show that the proposed model can improve the image visual effect effectively, and the adoptive numerical algorithm has faster convergence speed.关键词：image denoising;variation method;fractional-order gradient;saddle-point problem;primal-dual;staircase effect3515|254|0更新时间：2024-05-07
- 摘要：In order to evaluate different kinds of distorted images efficiently,a novel general-purpose blind/no-reference image quality assessment is proposed,which combines perceptual features with spatial natural statistics features to construct an image quality assessment model. Four perceptual features-phase congruency entropy,mean phase congruency,mean gradient, and entropy of the distorted images are selected beside the 36 spatial natural statistics features of sharp patches.features.Support Vector Machine Regression(SVR)is adopted to build the relationship between image features and quality scores,yielding a measure of image quality. Experimental results in the LIVE database show that the proposed method accords closely with human subjective judgment.It has good robustness and short running time. The proposed method has a good performance.The selected features are rational and the learning method is effective.关键词：blind/no-reference image quality assessment;perceptual feature;statistics feature;support vector machine regression3988|316|0更新时间：2024-05-07
- 摘要：At present, the algorithm of image scrambling degree evaluation has two problems. One is the instability of the evaluation result for the same scrambling algorithm, the other is the impossibility to evaluate many scrambling algorithms at the same time. In order to solve these two problems, a new evaluation algorithm for image scrambling is proposed based on analytic hierarchy process. At first, uniformity and randomness of scrambling matrix element distribution is calculated, thus, attribution values of scrambling pixels position distribution are acquired and the stability of the scrambling degree is guaranteed. Then, by applying analytic hierarchy process, the integration of two-dimensional attribution of scrambling degree and evaluation at the same time for many scrambling algorithms are implemented. Experiment results show that the scrambling degree of a scrambling algorithm is stabe and valid, it only relates with the algorithm itself and the number of iterations of the scrambling, and has nothing to do with the scrambling object. Meanwhile, the sorting of scrambling degrees for any number of scrambling algorithms is also established. The proposed algorithm is objective, fair and efficiency, it not only favors fast choosing suitable scrambling algorithm for the designer in information hiding field, but also provides powerful feedback tool for the development of scrambling algorithm.关键词：image scrambling evaluation;analytic hierarchy process;scrambling matrix;ideal uniformity distribution2891|298|0更新时间：2024-05-07
- 摘要：An effective object tracking method by combining the predicted target position with compressive tracking (CPCT) is proposed. Sparse Toeplitz projection matrices with random pitches are used to extract the compression feature of the original multi-scale Haar-like feature. Then, the distance between the sample positions and the predicted positions of the Mean Shift algorithm is used as object candidates' weights and the weighted Bayes classifier is used to determine the reliable object position. An adaptive parameter updating approach is used in the classifier training. The experimental results have shown that the average success rate of the CPCT tracking algorithm is about 27% higher than that of compressive tracking(CT), and the tracking speed is about 0.15 second per frame on average on 20 challenging sequences. CPCT tracking algorithm is more robust and smoother compared with six state-of-the-art algorithms on 20 challenging sequences by combining the predicted target position with compressive tracking and adaptive parameter updating approach.关键词：object tracking;projection matrices;compressive tracking;parameter adaptive3816|378|0更新时间：2024-05-07
- 摘要：To solve the problem of inaccurate segmentation and low character recognition on the instrument labels of small character spacing by traditional methods, a recognition algorithm is proposed for merged characters through adaptive segmentation and reconstruction. First, we use algorithms, such as median filtering and binarization, for image pre-processing. Second, the morphology filter is used to remove unwanted messages and to increase character spacing. Third, we find the geometric center of each character through centroid algorithm, getting each character bounding by the Sobel edge detection operator, and establish the Region of Interest (ROI)of each character. Then turn to the original image for character segmentation, and add certain pixels wide rectangular spacing bar, which refers to the GB standards to each character after segmentation, and then we reconstruct the image. Finally, we carry on the Optical Character Recognition (OCR). The character recognition experiments for 993 instrument labels with a recognition rate of 95.7%, The algorithm proposed was proved for an effective character recognition method for instrument labels.关键词：label character recognition;characters segmentation;mathematical morphology;image reconstruction;centroid algorithms;edge detection3506|428|0更新时间：2024-05-07
- 摘要：The detection of moving objects plays an important role in many image processing and computer vision applications, such as object recognition and tracking, as well as traffic surveillance. However, moving cast shadows are usually misclassified as moving objects because the moving shadows not only have the same movement as the moving objects but also have different intensity values with the background regions. It is well known that chromaticity and texture are the two most commonly used features in moving shadow detection. In this paper, by combining these two features reasonably, we propose a novel approach based on chromaticity and texture invariance to remove the shadow regions from the moving detection results. In our method, the moving shadows are detected not only at the pixel level, but also at the region level. When a background pixel is covered by shadow, its intensity value decreases, but the chromaticity remains the same. Therefore, we first use the chromaticity invariance at the pixel level to obtain candidate shadow regions. Chromaticity is usually used in HSV color space, which can provide a natural separation between chromaticity and luminosity. In our method, an approach applying chromaticity directly in the RGB color space is proposed, which allows for an algorithm withfewer parameters. In order to ensure that all the real shadows are included in the initial candidate shadow regions, a corresponding threshold is set conservatively in this step, which will lead to parts of the moving objects wrongly detected as shadow regions as well. To remove them from the candidate shadow regions, we first segment the candidate shadow regions into several sub-regions according to the color information. Since the moving objects and their shadows usually have different intensity values, this strategy can effectively separate the shadows from the moving objects. Then we introduce a quantitative index inspired by the local binary pattern (LBP) to measure the texture invariance of each sub-region. As regions under shadow tend to remain their textural characteristics, the real shadow regions can be finally detected in the region level. We choose five videos from a public dataset, which is most commonly used in moving shadow detection, to verify the effectiveness of the proposed method. The selected videos contain different scenes, indoor and outdoor, and the size of the shadow changes in a wide range. First, experiments to confirm the assumptions in this paper are conducted. Then, we compare our results with some classical and popular approaches via several objective evaluation criteria. The methods used for the comparison are chromaticity method, geometry method, physical method, small-region texture method, and large-region texture method. Experimental results demonstrate that our method can obtain good detection results and outperform other methods in terms of subjective visual perception and objective evaluation criteria. There are three main contributions of this paper. First, we present an approach, which directly applies the chromaticity invariance in RGB color space, which makes the number of parameters less than traditional methods using the HSV color space. Second, a quantitative index inspired by the LBP is designed to measure the texture similarity between the detected candidate regions and the corresponding background regions. Compared with conventional methods that use histogram-matching approaches, the proposed method is more accurate. Third, we provide a good detection example by combining the pixel level-based method and region level-based method. In this method, color information and texture information are both fully exploited. Experimental results demonstrate that our method can successfully detect the moving shadows in various categories of scenes.关键词：moving object detection;moving shadow detection;chromaticity invariance;texture invariance;local binary pattern3980|741|0更新时间：2024-05-07
- 摘要：In order to apply the manifold learning approach to image dimension reduction and recognition, an affine invariant adaptive locally linear embedding algorithm is proposed. Tangent distance is introduced and combined with locally linear embedding. In the sample space, the distance is described by an affine invariant image similarity based on the tangent distance method. The neighborhood size of every point in sample space is computed adaptively by simi-larity function. Experimental results show that the proposed algorithm is able to create low dimensional manifold structure more reasonably, and improve the recognition rate. The proposed algorithm is insensitive to affine transformation and performs more robust.关键词：Manifold learning;locally linear embedding;adaptive;affine invariant;tangent distance3226|231|0更新时间：2024-05-07
- 摘要：Human action recognition is a widely studied area in computer vision and machine learning; it has many potential applications including human computer interfaces, video surveillance, and health care. In the past decade, extensive research efforts focused on recognizing human action from monocular video sequences. Since human motion is articulated, capturing human joint characters accurately from video is a very difficult task. The recent introduction of real time depth cameras such as the Kinect sensor, give us the opportunity to use 3D depth data of a scene instead of pictures. In this paper, we present a manifold-based framework for human action recognition using depth image data captured from depth camera. With the recent release of Kinect sensor and the technology assessing skeleton joint position from depth image matured, recent research used 3D skeleton joint position information as human body representation and achieved good recognition performance. As we know, human action is composed of ordered posture set, and the difference between postures is only a few changes of 3D joints pairwise, most of the 3D information changes only little. In this paper, we estimated the 3D joint locations from Kinect depth images and use pairwise relative positions as the representation of human features. In the training phase, the LE (Lalpacian eigenmaps) is used to build action model in low dimensional space. In test phase, the nearest-neighbor interpolation technique is used to map test sequence to the manifold space, then measure the distance with the test sequence and the training data. A novel modified Hausdorff distance is used to measure similarity and fitness of the test sequence and the training data in the matching process. The recognition performance of the proposed method was evaluated from Kinect sensor dataset and the result confirmed the proposed method can work well in several experiments. We also tested the proposed method on the MSR Action3D dataset and achieved state of the art accuracy in our comparison with related work when the training set has many samples. Manifold learning is an effective nonlinear dimensionality reduction method and low-dimensional motion models can be trained well when training sample size is large. We propose a novel human action recognition based on manifold learning in this paper. The experimental results show the effectiveness of the proposed method for human action recognition based on depth image sequence.关键词：Kinect sensor;human action recognition;Manifold learning;Hausdorff distance;depth data4522|344|0更新时间：2024-05-07
- 摘要：To deal with the modeling limitation in traditional morphable models, we present a compressed sensing based method for 3D face reconstruction. In the proposed framework, compressed sensing is first used to estimate the similarity between testing and prototype face samples, and a modified morphable model is then built on the selected prototype samples with larger similarity. Second, the model is registered to a test face image by using facial salient points. Finally, combining the shape recovered by the modified model and the shape obtained by RBFs interpolation, the testing shape is reconstructed. Experiments on the BJTU-3D face database and CAS-PEAL 2D face database show that the proposed method outperforms the traditional methods in both reconstruction accuracy and computation complexity. Moreover, its recovered 3D faces achieve higher realistic impressions. This paper presents a novel fast 3D face reconstruction method. The proposed modeling plan, based on selected prototype samples using compressed sensing, can effectively improve modeling precision and efficiency. Combined with the RBF smooth scheme, the proposed method can reconstruct face shape with high smoothness.关键词：3D face reconstruction;morphable model;compressed sensing;radial basis function3789|460|0更新时间：2024-05-07
- 摘要：In order to enhance the performance of compressive sensing based tracking in complex scenarios, we propose an improved tracking algorithm based on a new feature selection approach and target model updating mechanism. First, we select features allowing to distinguish a target from the background, according to the distance between a feature's positive and negative class conditional probability Gaussian distributions. Second, we adaptively update the target appearance model according to the difference between the current target and the original one, so that the target would not be updated in case of big occlusion or frequent posture changes. Experiments on ten standard and complex test video sequences demonstrated that for the three measurements, i.e. center error, the success rate, and the precision plot, our algorithm, with the rate of 85.19% of frames correctly tracked and average 13.74 pixels of center location difference, achieves a higher perform than three state-of-the-art methods. The proposed new method of feature selection and target model updating, enhances the robustness of compressive sensing based tracking and speed up of the track.关键词：compressive sensing;object tracking;feature selection;target model update3582|396|0更新时间：2024-05-07
- 摘要：Since cameras are not able to capture a very wide range of brightness, the phenomenon of overexposure or underexposure is widespread, especially for high dynamic range (HDR) scenes. Image fusion technology aims to solve this problem. However, most of image fusion algorithms are too complex and time-consuming, Therefore, they are only suitable for processing static images. In order to solve this problem, many methods are used to optimize the original image fusion algorithm, which is introduced to deal with the exposure problems of video. We test the image fusion algorithms and make practical optimization for block-based image fusion to accelerate the original algorithm. Furthermore, the optimized algorithm is embedded in real-time monitoring system. Through combining the exposure control module with fusion technology, we design a video capture system with real-time image fusion. As it is illustrated by experiments in HDR scenarios, each frame of the video can maintain all the information of the scene without exposure problems. In this paper, many strategies for optimizing and accelerating the image fusion are proposed and the improved image fusion method is utilized to the real-time video capture ingeniously. The experimental results show that almost all the details are remained without exposure errors, even in HDR scenarios.关键词：image fusion;high dynamic range (HDR) scenario;real-time monitoring;exposure problems;optimization3103|348|0更新时间：2024-05-07
- 摘要：Simulation of Chinese ink is an important research topic in the non-photorealistic rending simulation field. The results of current rendering algorithms are stiff and colorless because the algorithms lack coherent edges drawing and the color convergence is not smooth. This paper presents a method converting real images into Chinese ink paintings. Firstly, this method utilizes an improved flow-based difference of Gaussian filter to generate edge contours. Then the anisotropic Kuwahara filter is utilized to converge the ink and make a natural transition. Finally, Perlin noise is introduced to simulate the plicate rice paper. The experimental result shows that the algorithm can achieve ink images using real-image as input and its processing speed is fast. This paper presents a method converting real images into Chinese ink paintings. The experimental result shows that the algorithms can achieve better effect images without destroying the original mapping structure.关键词：non-real imaging;image rendering;difference of Gaussian filter;Kuwahara filter;Perlin noise3671|462|0更新时间：2024-05-07
- 摘要：The precondition of a 3D model retrieval and synthesis with semantics is accurate segmentation and labeling. For the 3D character model widely used in games and movies, we propose a method of character parts segmentation and labeling for animation based on a priori knowledge of the structure. After the structure rules of the animation character model are generalized, the rules are formalized with a hierarchical knowledge expression that can instruct segmentation. The regional growth method is used to detach the model parts layer by layer when the extremity detection and the boundary judgment can be guided by the rules. Simultaneously the parts are labeled a semantic name. Our method is tested on the biped/quadruped character model through the structure rules description. The segmentation results are character components that are suitable for synthesis. Compared to other methods of character model segmentation with labeling, the result of our tests indicate that our method has more obvious parts, which can support model synthesis and has good extensibility to animation character.关键词：animation character;hierarchical structure;3D model segmentation;3D model labeling3032|291|0更新时间：2024-05-07
- 摘要：Tropical cyclones pose a serious threat to the economy as well as people's life and property in the southeast coastal areas of China. Stationary satellite imagery is the main data source for tropical cyclone real-time monitoring. On the satellite cloud image, the textures of tropical cyclones are similar to that of other cloud structures, which makes the automatic extraction of tropical cyclones difficult. Base on a vector square, a concept of rotation coefficient is proposed to describe the characteristics of tropical cyclones. Additionlly, a tropical cyclone auto-recognition method is also presented. The Otsu algorithm is used to obtain the segmentation threshold, then rotation coefficient combined with cyclone area and brightness temperature features are implied to recognize tropical cyclone. The contrast experiment of the original vector square method and the improved method was carried on, using the imagery of typhoon HAIKUI. The statistics results generated throughout the life cycle stage shows that, the recognition rate of the improved method are 76%, 95% and 78% respectively, which are higher than that of the original method. Experiments show that relative to the vector square algorithm, the segmentations of tropical cyclone are more complete and the recognition rate of tropical cyclones in different development stages is higher.关键词：tropical cyclone;FY-2E;Otsu;rotation coefficient;automatic recognition3573|282|0更新时间：2024-05-07
- 摘要：Fusion strategies based on a nonsubsampled contourlet transform (NSCT) are able to restrain the background information and enhance the information of changed regions in the fused difference image effectively. However the fused difference image has complex statistical characteristics, the conventional change detection based on statistical characteristics is difficult to achieve. An unsupervised detection for flood change algorithm in SAR image based on NSCT fusion and parametric kernel graph cuts is proposed in this paper. The difference images acquired by mean-ratio operation and log-ratio operation are integrated by the NSCT fusion algorithm, the fused difference image is divided into foreground and background to obtain the final change detection results by the parametric kernel graph cuts algorithm. Change detection accuracy of the fused difference image is improved by taking advantage of complementary information from mean-ratio and log-ratio images. The approach with strong applicability is distribution free, and does not need any prior knowledge. Experimental results show that the detection accuracy of the proposed algorithm is superior to the traditional change detection algorithm.关键词：nonsabsampled Contourlet transform(NSCT)fusion;parametric kernel;graph cuts;change detection;SAR image2865|323|0更新时间：2024-05-07
- 摘要：In many cases, the cone-based model and the direction-relation matrix model may make mistakes in determining the direction relations between two objects, owing to not properly taking into account the impacts of factors on the direction relations between two objects, such as the shape and size of the objects, as well as the distance between them. For this reason, a compound model is proposed by combining the cone-based model and the direction-relation matrix model to overcome the deficiencies of both to describe well the qualitative direction relations between two objects. The basic idea is: first, redividing all direction regions by means of the geometric operations between the cone-shaped direction regions of the cone-based model and the corresponding rectangular direction regions of the direction-relation matrix model to construct the compound model;then, getting the qualitative direction relations from the reference object to the target object by computing the intersection of the target object and each direction region of the compound model and using the matrix to store the results. The experiments show that the compound model avoids effectively the defects existing in the cone-based model and the direction-relation matrix model. Additionally, the compound model can accurately describe the qualitative direction relations between two objects and provide supports for spatial reasoning and spatial queries the defects existing in the cone-based model and the direction-relation matrix model.关键词：cone-based model;direction-relation matrix model;direction relations;compound model2784|259|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