最新刊期
卷 18 ， 期 3 ， 2013
- 摘要：In order to reduce detection error in high density salt and pepper noise and to improve the filtering performance, a new local adaptive gray similarity bilateral filter algorithm is proposed in this paper. First, the algorithm extracts quasi-noise through gray-scale extremism of salt and pepper gray values, and then detects the local window border for every quasi-noise point to get the possible noise collection. Finally, only the pixels in the quasi-noise collection need to be filtered with nearby bilateral weighted gray-scale filtering. Experimental results demonstrate that the new algorithm has increased the parameters for peak signal-to-noise ratio(PSNR) by 0.2~1.6 dB in different noise density. Experimental results demonstrate that the new algorithm has the most integrated optimal performance in comparison to other filtering algorithms.关键词：gray-scale difference;based on local border;principle of the nearest distance;salt and pepper noise3136|444|0更新时间：2024-05-07
- 摘要：In order to solve the problem of the curse of dimensionality and the small sample problem, a kernel sparsity preserving projection is proposed. First, the nonlinear transformation is used to map the original data to a high-dimensional feature space. Then, the sparsity reconstruction in a high-dimensional space is used and, the coefficient matrix is reduced and optimized. Finally, the projection matrix is obtained. This method is evaluated on the CASIA (B) Gait database. The experimental results show that the proposed method can obtain stable classification and performs satisfactory recognition results.关键词：sparsity preserving projections;kernel method;gait recognition;gait energy image3031|271|0更新时间：2024-05-07
- 摘要：Due to the coarse fineness of global Gist features in scene categorization tasks, we propose a local Gist feature description based on a dense grid. It uses a spatial pyramid structure to add distribution information and introduces the RGB color space to add color information. The feature matching process is kernelized by an efficient match kernel which mea-sures the similarity between local features based on the BOW model. The scene categorization task can be done with linear SVM. Experiment shows the influence to the classification accuracy with local Gist features which have different scale, orientation, fineness, match kernels and numbers of training samples. By using the classification result of the global Gist feature and dense SIFT features on the OT scene dataset, we demonstrate that the proposed feature construction method and classification model are efficient.关键词：local Gist feature;spatial pyramid;efficient match kernel;scene classification6075|303|0更新时间：2024-05-07
- 摘要：A method of scene description based on the conditional random field model is presented in this paper. The conditional random fields models the posterior directly, so that it can exploit several types of features, and has the ability to contact context information. Therefore, the CRF model in the scene description can get a more accurate description of the results. In this paper, the images are divided into rectangular blocks with a size of ×. The color feature, texture feature, and location feature for each rectangular block are extracted through multi-class features extraction. These features are clustered by the K-means algorithm, and then the feature vector is composed of the features clustered by K-means in accordance with the position of the rectangle. The feature vector is modeled by the CRF model. The model parameters are estimated through training. We use the MPM algorithm for model inference to get the scene description. The experimental results show a higher accuraly of the method presented in this paper for scene description.关键词：scene description;feature extraction;K-means;conditional random fields3347|262|0更新时间：2024-05-07
- 摘要：In this paper, we present a new technique for solving geometric structure and motion problems based on L and scene point classification. First, in order to avoid that our reconstruction method is affected by outliers, the outliers are removed using the L approach. Second, the scene points are divided into two classes; the first class contains those scene points which are visible in two views; the second class contains those scene points which are visible in more than two views. The optimal triangulation method is used for the first class. The L norm minimization is used for the second class and the gap between the upper and lower bound of the bisection is kept as small as possible, so that the computation time is decreased. The experimental results show the advantages of our method.关键词：3D scene reconstruction;scene point classification;second order cone programming;L;approach;norm;the optimal triangulation method2531|275|0更新时间：2024-05-07
- 摘要：In this paper, we propose a new object tracking algorithm applying sparse representation in the Lucas-Kanade image registration algorithm. The object state parameters are solved to realize precise tracking by minimizing the L-norm of the alignment error. The object appearance is represented by the static template and the dynamic dictionary. The dynamic dictionary is obtained by updating the tracking result in each frame. The object can be rebuilt by the sparse representation of the templates in the dynamic dictionary. To deal with tracking drift caused by dictionary update, a two-stage iteration with the static template and the dynamic dictionary respectively is included in our method. Numerous experimental results show that the proposed method is quite effective to partial occlusions, appearance changes and illumination changes. Meanwhile the system is computational efficient and works in real time.关键词：sparse representation;Lucas-Kanade image registration algorithm;visual tracking;L;-minimization4394|416|0更新时间：2024-05-07
- 摘要：In this paper, we propose a nevo model for three dimensional vessel extraction. The model makes full use of tubular properties, which includes prior intensity of vessels, second tensor of tubular structures, and geometric curvatures. All this information makes the energy terms of the active contour model and thus leads to three forces of the iterative equation: the region competition force using prior intensity, the tubular vector field force and the dual curvature force. The first force help extracting big vessels accurately and robustly, the second force makes it possible to extract thin and weak vessels, and the last one is able to remove noise without changing the tubular geometry. As shown in the experiments extracting liver vessels, coronary, and lung vessels, the proposed model is able to extract the whole vessel trees automatically, accurately, and robustly.关键词：vessel;segmentation;Hessian matrix;active contour model;liver;Coronary4060|300|0更新时间：2024-05-07
- 摘要：In the cyber security visualization of field, the multi-step attacks visualization has shortage in interacting with logs. This cannot be effective to make the network security administrator find multi-step attacks using logs. In this paper, we present a multi-step attack visualization tool, which is based on rules tree. It describes multi-stage attacks model by rules tree, defines a template library by XML and designs visualization models. It represents the multi-step attacks scene by comparing vector visualization and three-dimensional visualization, and finds the advantages. Our experiments prove the tool’s validity and the design’s rationality.关键词：security visualization;rules tree;multi-step attacks;alert correlation;visualization representation4508|199|0更新时间：2024-05-07
- 摘要：Considering the shortage of salient edge information for existing image abstraction methods, we present a salient edge guided approach based on energy optimization. For an input image, we first construct a salient edge map by a proposed edge information passing scheme which can effectively reduce the discontinuity of long edges. Then, in order to accentuate the details of the salient edges while suppressing the tangle-some details, we build an expected image gradient field in terms of the salient edge map. Finally, constrained by the image color information and the expected gradient field, we get the rendering result by minimizing a devised energy function. The experiment results show that our method has obvious advantage in the preservation of the edge coherence.关键词：image abstraction;salient edge;expected gradient field;energy optimization4563|302|0更新时间：2024-05-07
- 摘要：As motion capture data is widely used nowadays, the compression of motion data becomes more and more important. In this paper, a sparse representation based approach is proposed for efficient compression of human motion data. A new algorithm is designed to extract the dictionary from an input motion clip automatically. Each frame of a motion clip can be represented by a sparse linear combination of the dictionary vectors. The experimental results show that our method can get a high compression ratio (about 50 times) for general short motion data, with a limited reconstruction error, which is hard to visually distinguish (ARMS error less than 2.0).关键词：motion data compression;sparse representation;dictionary;compression ratio;overall error4357|231|0更新时间：2024-05-07
- 摘要：Rendering based on user interests is one of the hot topics in the expressive rendering of large-scale scene. Objects with high user interests should be rendered in detail, while coarse rendering does well for objects with low user interests. Traditional methods compute objects’ user interests based on their spatial distances with the user’s central focus object. In this paper, we propose a novel semantic-based approach for computing objects’ user interests. In this approach, we first construct the scene’s semantic forest model and compute the semantic distances between any two semantics in the preprocessing stage. At runtime, after the user specifies the central focus object, with its semantic and spatial information we can calculate the user’s objects of interest in the scene considering both, their semantic and spatial distances within the user’s central focus object. Finally, the objects are classified by their user interests and rendered with multiple styles. The user interests calculated by our method conform to human perception and our method can achieve real-time performance.关键词：semantic tree;user interests;scene classification;expressive rendering4224|284|0更新时间：2024-05-07
- 摘要：Books image retrieval is an important application of CBIR (content-based image retrieval). SIFT(scale inva-riant feature transform) and HOG(histogram of oriented gradient) are widely used in CBIR and object recognition. A book image is divided into nine blocks called sub-images according to its characteristics. We extract features from the image based on SIFT, and then we combine the features with the HOG feature to create a combined feature. The combined feature occupies little data and describes an image well. We use a new way to calculate the combined features’ similarity, which will avoid large errors caused by some sub-image. An image set with 50000 book images is used to test our method and we find the accuracy to be acceptable and the time consumed by the retrieval process is quite short.关键词：books image retrieval;scale-invariant feature transform;histogram of oriented gradient;grid space frame4474|257|0更新时间：2024-05-07
- 摘要：A new three-dimensional segmentation approach of retina Optical Coherence Tomography (OCT) volume data is presented, which makes full use of the retina layers’ directions, signal intensity peaks and divide-and-conquer strategy to detect multiple retina boundary surfaces. The primary position of the inner limiting membrane (ILM) is determined by finding the first peak in every A-scan from top to bottom in the volume. Abnormal points of the boundary are removed by using surface smoothness constraints. The retinal pigment epithelium (RPE) boundary surface is obtained similarly by finding the first peak in every A-scan from bottom to top. Then, the boundary surface of the inner segments and outer segments (IS/OS) layer is positioned by detecting the maximum intensity of every A-scan in the sub-volume data between the ILM and RPE. Finally, the boundary surface of the inner nuclear layer and the outer plexiform (INL/OPL) layer is segmented by searching the first peak of every A-scan from bottom to top in the sub-volume data between the ILM and IS/OS. Experimental results show that the method can not only segment the four boundary surfaces correctly, but also only takes several seconds for segmenting each surface on a personal computer.关键词：optical coherence tomography;retina;volume data;three-dimensional segmentation4794|232|0更新时间：2024-05-07
- 摘要：Real-time soft shadow algorithms can render convincing soft shadows on complex and dynamic scenes with a single shadow map. In this paper, we propose a hybrid rendering method, which extracts edges for area light in object space and determines the edge pixel’s direction in image space. Then, we use the Hierarchical Edge Map to compute the visibility. The results show, the cost of silhouette extracting is reduced and the rendering performance is improved. Furthermore, the accuracy of the contour and the shadow quality are both enhanced.关键词：real-time soft shadow;area light;object space;edge extraction;backprojection4231|362|0更新时间：2024-05-07
- 摘要：Crowd simulation has been widely used in industry, architecture, transportation, and many other fields. To implement real-time crowd simulation in complex environments, efficiency is a pivotal problem to resolve. We meet a lot of challenges, such as rendering of large crowds, the update of crowd’s locations and states, as well as collision avoidance. We propose a field-based approach to implement real-time crowd simulation. This approach guides the crowd movements by constructing a navigation field and a density field. The navigation field can make the crowd choose the optimal path to reach the desired destinations. The density field can affect the velocity of the crowd to help avoid collisions, combined with a GPU-based collision avoidance method. Using our approach, we have constructed a real-time crowd simulation system and tested the performance in a large venue with thousands of agents. We succeed in simulating the evacuation of crowds with excellent rendering and high efficiency.关键词：crowd simulation;path planning;collision avoidance;crowd motion;density field4735|266|0更新时间：2024-05-07
- 摘要：In order to achieve accurate automatic image reparation, we propose a new framework for image completion based on the large displacement view(LDV)image. First, we extract the evenly distributed and quasi-dense feature-point correspondences between the target image and the LDV image by combining multiple distinct feature detectors in a complementary way. Inspired by the prior model and model fitting problem, we then devise a quasi-planar scene-regions(QPSRs)clustering algorithm, which classifies the feature point correspondences and represents a natural scene image with multiple QPSRs to remove the perspective distortions in the LDV image. Inspired by the texture synthesis and image stitching techniques, we finally present a QPSRs compositing algorithm, which corrects and stitches the re-projected QPSRs images to fill in the missing areas on the target image. Our experimental results are comparable with the ground-truth.关键词：image completion;large displacement view;quasi-planar scene regions;clustering;compositing4305|482|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