最新刊期
卷 16 ， 期 2 ， 2011
- 摘要：Local invariant features are receiving increasing attention from computer vision research community. Local invariant features have been widely utilized in a large number of applications, e.g., wide baseline matching, object recognition, and categorization, image retrieval, visual search, robot localization, scene classification, texture recognition and data mining. This paper gives an overview of the various approaches and properties of local invariant features. We focus on three major areas: (1) local invariant feature detectors, (2) local invariant feature descriptors, and (3) local invariant feature matching. Most of the existing local invariant feature detectors can be categorized into corner detectors, blob detectors or region detectors. Local descriptors can be categorized into distribution-based, filter-based, moment-based descriptors and others descriptors. Similarity measurement, matching strategy and matching verification are three key components of robust matching algorithms. Finally, some research challenges and future directions are discussed.关键词：local invariant features;local invariant feature detector;local invariant feature descriptors;local invariant feature matching5111|480|0更新时间：2024-05-07
- 摘要：In recent years, since data acquisition equipments have been improved, a growing number of high-precision large images appear in the medical imaging, remote sensing, meteorology, archeology and other fields. These images not only provide users with detailed information, but also bring out tremendous challenge to display them interactively. We designed a scalable parallel pipeline, according to our algorithm based on proposed method of data management. Then, we design the algorithm to zoom the large image based on CUDA(compute unified device architecture) framework interactively. Our algorithm does not require any pre-processing operation, can obtain excellent real-time performance and stability, and improve the efficiency of the browser greatly.关键词：multi-core;large image;real-time4020|235|0更新时间：2024-05-07
- 摘要：Enhance edges of images efficiently and removal noises in images are fundamental issues in the field of image processing and computer vision. A rescaled power space filtering method is proposed in this paper using multi-scale dyadic wavelet transform for 2D noisy image. Some sample images are tested and results demonstrate that the rescaled power space filtering method can preserve edges features and reduces noises. Compared with the commonly-used wavelet hard-thresholding and soft-thresholding methods, rescaled power space filtering algorithm increases the peak signal to noise ratio (PSNR) by 2~3dB which indicates the superior denoising performance of the power space filtering method.关键词：dyadic wavelet transform;rescaled factor;peak signal to noise ratio3773|346|0更新时间：2024-05-07
- 摘要：Shearlet is a new-style multi-scale geometry analysis tool. It creates Shearlet functions which have different characteristics through zooming, shearing, translating and other affine transforming methods, and enables its capable of optimally sparse representation. Firstly, a digital Shearlet transform implementation method is proposed in this paper. And then, a new de-noising method that combines Shearlet transform and variation is presented, which mainly been established using a total variation regularization model to constraint condition on Shearlet transform domain. The proposed model aims at reducing Pseudo-Gibbs artifacts after simple threshold methods. Numerical examples demonstrate that the method can remove noises and keep edges effectively, leading to improved visual effect and PSNR.关键词：shearlet transform;total variation;image de-noising;PSNR4392|240|0更新时间：2024-05-07
- 摘要：The microcalcification information is an important foundation for the diagnosis of breast cancers. In order to improve the badly vision of microcalcifications which are affected by noises and tissues, a novel enhancement algorithm of mammograms based on shape-selective filter and adaptive background suppression is presented. First the potential non-linear microcalcifications are detected by using shape-selective filter, and then the mammograms are divided into foreground and background sections. Second the background information is suppressed by using an adaptive contrast suppression method, and the foreground information is enhanced at the same time, then the key information of the mammograms is enhanced selectively. Experimental results show that this algorithm can enhance the microcalcification images and suppress the influence of blood vessels and tissues.关键词：image enhancement;shape-selective filter;background suppression;mammograms2945|197|0更新时间：2024-05-07
- 摘要：PROPELLER(periodically rotated overlapping parallel lines with enhanced reconstruction) is a new acquisition technique which can efficiently reduce motion artifacts in MRI imaging. Convolution gridding method usually necessitates lots of parameters optimization and a sampling density compensation step, so the quality of the reconstructed image cannot be ensured. In the paper, an iterative method is applied to reconstruct images for PROPELLER MRI. In the method, a cost function is iteratively minimized by using weighted pre-conditioned conjugate gradient algorithm. In order to improve computation, NUFFT(nonuniform fast Fourier transformation) is used to computing matrix-vector multiplication. Experimental comparison was made by using both digital phantom data and experimental PROPELLER imaging data. The results showed that the iterative method can improve signal to noise ratio of images and reduce ring artifacts of images in comparison with convolution gridding method. The homogeneity of images can be improved as well.关键词：magnetic resonance imaging;convolution gridding;PROPELLER;iterative method3061|180|0更新时间：2024-05-07
- 摘要：A modified adaptive Kalman filtering algorithm considering the system matrix uncertainty and data instability of the state space theory for Positron emission tomography(PET) reconstruction was proposed. Based on tracer kinetic theory, an evolution equation of the tracer is introduced as a prior to constrain the reconstruction. Along with the observation equation of the detectors, the two equations constitute a state spatial model. After introducing virtual noise to represent the error of the system matrix, the modified adaptive Kalman filter is applied to estimate the process and the observation noise and meantime completes the PET reconstruction. The performance of the algorithm was verified by computer simulations, which show that modified adaptive Kalman filter is more robust than the traditional maximum likelihood expectation maximization method and filtered back projection methods. The results are meaningful and particularly suitable for the real positron emission tomography system.关键词：system matrix uncertainty;state space;adaptive Kalman filtering;virtual noise;robust3258|226|0更新时间：2024-05-07
- 摘要：Frame rate up conversion (FRUC) is a conversion process between any two display formats with different frame rates. It can be used in HDTV application and in low bandwidth video coding. In this paper, a scene change detection method is proposed to detect whether the scene has changed. If the scene change occurs, the extrapolation technique will be used, otherwise, the interpolation will be chosen. Second，a motion vector processing method is adopted to correct and refine the unreliable motion vectors with variable block sizes, so the visual artifacts due to incorrect motion vectors can be eliminated and the object boundaries and structure information can be preserved. Experimental results show that the proposed algorithm provides good subjective and objective quality.关键词：frame rate up conversion;motion compensation frame interpolation;scene change detection;motion vector processing4002|190|0更新时间：2024-05-07
- 摘要：Blocking artifacts are serious artifacts caused by block-based hybrid video coding, and significantly impair the subjective quality of the compressed video. In this paper, the fundamental reasons of blocking artifacts are explored, based on which an effective quantization method for DCT coefficients for reducing blocking artifacts is proposed. The proposed method adjusts the quantization process to reduce the variation of the quantization error at block edges, which increases the correlation of the quantization error of adjacent blocks and therefore reduces the blocking artifacts of the compressed video. Experimental results have demonstrated the effectiveness of the proposed method in reducing blocking artifacts and improving the subjective quality of video frames. Furthermore, the proposed method is practical in implementation since it has a low computational complexity and is compatible with video coding standards.关键词：video coding;blocking artifacts;quantization3885|189|0更新时间：2024-05-07
- 摘要：Variable Length Coding is one of the key technologies in video coding with large computation requirement. A high speed and efficient VLSI architecture for H.264/AVC CAVLC is proposed. A new look-up table algorithm based on the pre-processing greatly optimizes look-up table structure. An arithmetic compute method is exploited to replace look-up table, and the arithmetic expression is optimized with technology of unfolding and share. Some area optimization is done in the other part. Experimental result shows that for logic synthesis, the hardware cost of the proposed design is 8723 logic gates by using SMIC 0.18μm CMOS technology at the clock frequency constraint of 133MHz. The new architecture can meet the requirement for the real-time processing for High-definition 1920×1088-30fps video with less hardware cost. So it has practical application value.关键词：H.264/AVC;context-based adaptive variable length coding(CAVLC);VLSI architecture3153|209|0更新时间：2024-05-07
- 摘要：It is necessary to establish a benchmark for information hiding. Imperceptibility is an important index to evaluate the performance of steganography. Based on HVS model，a new imperceptibility measure CHD for color stego image was proposed．Experimental results show that the quality of stego image calculated by CHD reflects the luminance masking, texture masking and color masking of HVS，and CHD can be used to evaluate the imperceptibility of steganography tools．关键词：information hiding;evaluation;HVS;imperceptibility;peak signal to noise ratio3951|257|0更新时间：2024-05-07
- 摘要：An adaptive fuzzy C-means image classification algorithm based on SSCL is proposed, in order to overcome the shortcomings that traditional fuzzy C-means clustering algorithm is noise-sensitive and relies excessively on initial cluster centers. First we obtain the cluster centers using SSCL, then treat the cluster centers as the initial value of fuzzy C-means, so an adaptive image classification can be achieved. At last, post processing is implemented using space information. Experiment results show that proposed algorithm is less sensitive to noise and initial cluster centers in FCM method, and has better classification accuracy.关键词：image segmentation;fuzzy C-means;self-splitting competitive learning3898|296|0更新时间：2024-05-07
- 摘要：Segmenting natural image into regions is concerned as an essential issue in computer vision. In this paper, a new weighted graph-cut algorithm which uses HSI color model is presented to solve a natural disaster image segmentation problem. Using graph theory, the color image is firstly translated into weighted graph by mapping each pixel into graph node. Weighted function is defined by combining HSI color and pixel distance factors, and Cut criterion is suggested by using region internal differences and region external similarity. The experimental result is provided to show the effectiveness of the algorithm, and the comparison with other methods results are also given in detail.关键词：graph theory;image segmentation;HSI distance3854|299|0更新时间：2024-05-07
- 摘要：Breast region extraction of mammogram is a crucial step for further analysis and processing, and can improve diagnosis accuracy of galactophore diseases. An efficient algorithm of automatically extracting breast region is proposed. The discontinuity of the variety of areas of the isosurfaces is explored and used for computing the accurate segmentation threshold. Then the connected region containing the breast region is extracted by using a scan line based method, which is more efficient than seed-filling based methods. In order to obtain more pure breast region, the unexposed borders of the image connected to the breast region are also removed through a few refinement procedures. Experimental results show that the method has good performances both on the accuracy of breast region segmentation and executing speed.关键词：mammogram;breast region;region growth;thresholding segmentation;isosurface3202|191|0更新时间：2024-05-07
- 摘要：The traditional pixel based background model cannot represent the background motion efficiently. In this paper, a novel strategy is proposed to model background and track moving objects based on optical flow velocity field. Statistics on intensity, spatial and temporal information of pixels are extracted to generate the optical flow field, which is used to formulate a novel background model for tracking moving objects efficiently and exactly. This optical flow field based strategy can reduce noise generated by background motion significantly and track moving objects robustly, as illustrated in our experiments.关键词：background modeling;optical flow;object detecting;mixture of Gauss3282|311|0更新时间：2024-05-07
- 摘要：According to the characteristics of human visual system, a salient object detection method based on spectral residual and multi-resolution is proposed. We first compute the spectral residual of three features i.e. intensity, color and orientation under different scales to build a series of multi-resolution saliency maps, which can be combined through linear interpolation to generate three feature-saliency maps. Then we use k-means clustering for binary clustering and select the feature-saliency map with the largest distance between two centroids. Finally we apply dynamic threshold segmentation to get salient regions in an image. The experimental results on natural images show that the new algorithm is stable and practical, and we achieve satisfied results.关键词：visual attention;objects detection;spectral residual;multi-resolution analysis;k-means clustering4580|328|0更新时间：2024-05-07
- 摘要：A novel approach that based on fan projection histogram to obtain the lane of the structured road is proposed in the paper. The objectives of approach are to reduce the AOI of vehicle recognition, the demand of the vehicle status and the difficulty of vehicle recognition in the driving direction, as well as to strength the robustness of methods on vehicle recognition. Both models of choice appreciation of projection point and partition of regional angles of lanes are created. Results demonstrate that the method is suitable for both the straight lanes and lanes of moderate curvature. The method shows little dependence on the direction of the vehicle, which owns strong ability to resist the disturbances caused by the vehicle shades, intense sunshine and long serious lane absence. It can also resist the disturbance resulted from deceleration belt and adapt to stronger or weaker light to some extent.关键词：AOI;fan projection histogram;projection point;lane snatch3455|607|0更新时间：2024-05-07
- 摘要：Traditional Mean Shift (MS) algorithm can only follow objects with translation and scale change, and fails to handle objects with similarity transformation or complex affine transformation. To address this problem, the paper presents two improved algorithms. The first one focuses on the affine motion. According to the theory of Singular Value Decomposition, the affine matrix can be factored into product of two rotation matrixes and one diagonal matrix, based on which a new candidate model is proposed. With Bhattacharyya coefficient as a similarity function, the object tracking is formulated as an optimization problem, and the corresponding MS algorithm can be derived by calculating the first derivative of the similarity function with respect to affine parameters and setting them to be zero. Furthermore, a new candidate model is proposed that handles similarity transformation, and the corresponding MS algorithm can be obtained that estimates the translation vector and rotation angle. Experimental results show that, the proposed algorithms can track objects with similarity or affine tranformations, and have better tracking performance than the traditional one.关键词：object tracking;Mean Shift algorithm;affine transformation;similarity transformation3936|268|0更新时间：2024-05-07
- 摘要：For real-time tracking of moving vehicles, the general and efficient method is based on Kalman filter. However, the false detection often exists when more than one objects approaches with each other. It causes error in target matching process. To overcome the above problem, this paper first considers the width/height ratio and occupancy ratio to make false detection judgment. Then a new moving vehicle segmentation algorithm based on feature points on contour is presented. Lastly, the tracking model based on extended Kalman filter is implemented. Experimental results demonstrate that the rule can recognize false detection quite accurately. The proposed vehicles segmentation method can segment the overlapped ones accurately and completely, and finally the tracking model based on extended Kalman filter is implemented to realize the real-time tracking.关键词：object segmentation;feature point;extended Kalman filter;vehicle tracking3672|278|0更新时间：2024-05-07
- 摘要：We propose an approach for eye corner location based on weighted variance projection function(WVPF). The rough positions of eye corners are first detected using Haar like features by Adaboost. Then, the weighted variance projection is designed by assigning the corner response function of Harris corner detector as our weight factor. Owing to the existence of both peaks of eye corner position at the projection curves of WVPF in horizontal and vertical directions, the eye corner positions are located by analysis of peaks of WVPF. Testing results on BioID face database show that our approach obtains high accuracy and low computation complexity.关键词：eye corner;WVPF;corner response function;location3154|215|0更新时间：2024-05-07
- 摘要：There exists complementarity between different images of one scene. A better image can be obtained by fusing these images in gradient domain. The structure tensors of the images are fused and the local contrasts are incorporated in the design of fusion weights. The target gradient field whose structure tensor approximates the aforementioned tensor in the Frobenius norm sense is then obtained. An enhanced image is finally reconstructed from the target gradient field by least square fitting. Applications can include fusion of results by different enhancement methods, photos of the same scene with different focus or different exposures, etc. Experimental results demonstrate that the fused image can preserve significant details and structural information of each input image channel and the visual effect is improved.关键词：image enhancement;gradient domain;image fusion;structure tensor5027|294|0更新时间：2024-05-07
- 摘要：The estimation of intra frames homography is a key issue to image registration and mosaicing, and the conventional estimation methods are only feasible for the gray images. In this paper, the homographic matrix of two colored consecutive frames is proposed based on a hierarchical motion estimation approach. In our algorithm, the new constraints, viz., the conservation of hue and saturation, are adopted to obtain optical flow equations and the defects of the lightness conservation are remarkably decreased. Meanwhile, the numerical derivatives of the images are obtained by using the optimal multi dimensional derivative filters which enhance the robustness and accuracy of the algorithm. Furthermore, the scaled total least squares (STLS) approach instead of the least squares (LS) or total least squares (TLS) method is used to estimate the parameters of the homography and the adaptability for the image noise are enlarged greatly. The experiment results with synthetical and real images show that the algorithm is of high robustness and accuracy and the dense image corresponding points can be obtained as byproducts.关键词：color image;homography;optical flow;scaled total least squares4150|213|0更新时间：2024-05-07
- 摘要：A practical image alignment method based on self-adapting subdivision mesh is proposed, which can align the projector imagery on arbitrary smooth curved screen with camera image. It is designed for calibration problem of multi-projector displays. Using sparse initial corresponding points between the frame buffer of projector and the image of camera, the proposed method can subdivide it into arbitrary precision. Hence, the one to one correspondence of pixels between the frame buffer of projector and image of camera can be established. This correspondence can be used as the foundation of methods for the calibration problem of multi-projector displays. Comparative analysis of existing algorithms and real applications on virtual airport control tower simulation system show that the proposed method can align images more accurately, without any explicit prior knowledge of projector, camera and screen.关键词：imagery alignment;multi-projector displays;self-adapting subdivision mesh;projection on curved surface4902|244|0更新时间：2024-05-07
- 摘要：Perspective shape from shading can be solved by combining static Hamilton Jacobi(HJ) equation and fast sweeping method. To optimize the solution of static HJ equation and improve the results of perspective shape from shading,this paper uses high order local Lax Friedrichs(LLF) flux splitting scheme to increase the accuracy of partial derivatives.Besides, the paper uses modified weighted essentially non oscillatory(WENO) scheme, which only uses integer grids and gets higher accuracy than WENO by using amended smoothness estimators. Experiments of synthetic and real images demonstrate that the algorithm can obtain accurate results for perspective SFS.关键词：shape from shading;static Hamilton Jacobi equation;fast sweeping method;local Lax Friedrichs;weighted essentially non oscillatory2815|210|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