最新刊期
卷 16 ， 期 10 ， 2011
- 摘要：The quality of side information generated by a decoder has a significant influence on the Rate-Distortion (RD) performance of distributed video coding (DVC) codec. In order to improve the RD performance of a DVC by improving the quality of side information, a new side information generation algorithm based on pixel level motion vector field (SGPMVF) is proposed. First, the accurate pixel level motion vector fields (PMVF) of the forward and backward key frames are acquired respectively by motion estimation; then, the motion vectors of all pixels in the Wyner-Ziv frame are selected; finally, the high quality side information is obtained by interpolation. A new search range determination method based on the difference of SAD (sum of absolute difference) and a new pixel level MV (motion vector)selection method is proposed in the SGPMVF. Experimental results show that up to 1.45 dB gain for the side information and about 0.3～0.7 dB better RD performance can be achieved for fast motion sequences, while remaining the same rate and coding complexity.关键词：distributed video coding (DVC);side information generation;pixel level motion vector field (PMVF)3197|243|0更新时间：2024-05-08
- 摘要：In order to overcome the weak robustness，the weak reservation of edges, and the high sensitivity to noise in some super-resolution methods, we propose a nonlocal-means super-resolution reconstruction with enhanced edges based on the MAP frame. This method adopts the nonlocal-means model, and computes the modulus of the model together with the edges of the image. The proposed method mends effectively super-resolution reconstruction based on the bilateral total variance (BTV) model and based on the Markov random field (MRF) model. Our method is more robust and it is more able to reserve edges and to remove noise. Experimental results show that the proposed method is robust, and can reserves the edges well under low signal to noise ratio, getting a better reconstruction result.关键词：super-resolution reconstruction;MAP;non-local algorithm2952|220|0更新时间：2024-05-08
- 摘要：An algorithm for image reduction and zooming based on Hilbert scaning and the spring model is presented in this paper. Through the Hilbert scan, the two-dimensional image is converted to a one-dimensional gray-level sequence. Then, the gray-level sequence is imaged to a rubber band, and the image zoom and reduction can be realized by stretching out and drawing back the rubber band. Our experiments show that the algorithm is simple in operation, and provides good results.关键词：Hilbert scan;spring model;image reduction;image zoom2758|195|0更新时间：2024-05-08
- 摘要：For Bayesian dynamic Positron emission tomography (PET) reconstruction, a regional spatial-temporal prior (RST) model is proposed in this paper. The RST exploits sufficiently both the spatial and temporal information, and could suppress the noise in both two spatial dimensions and one temporal dimension. Computer simulations of brain 18F-FDG dynamic PET was conducted to validate the proposed approach. The comparison with other classical reconstruction methods shows that the proposed approach performs better in dynamic PET reconstruction and the results are more accurately estimating of the influx rate Ki of 18F-FDG in the lesion region.关键词：Bayesian reconstruction;dynamic PET;regional spatial-temporal prior;compartment model3602|197|0更新时间：2024-05-08
- 摘要：Boundary Structural irregularity of skin lesions has shown the most significance in the early diagnosis of melanomas in clinic. Aiming to measure structural irregularities of skin lesion contours and designing feature oriented classifiers, a method for extracting structural components of contours based on wavelet sub-band analysis and neural network classifiers using multi-scale features is proposed in this paper. A set of salient wavelet sub-bands are searched using Hausdorff distance evaluations on energy at wavelet subbands. Then structural components of skin lesion contours are reconstructed. Furthermore, multi-scale descriptions of the contours structural irregularity are given using statistical and geometrical measures based on the salient wavelet sub-bands. Finally, various experimental schemes are taken on single scale/multi-scale features, for small training sets and large training sets as well as and before and after feature selection for a comparison on performances of the neural network classifiers. It is shown from experiments that the classifiers sensitivity and specialty after feature selection using small sample sets equal and exceed that of large training sets respectively as dimensions of feature spaces are expanded by the proposed multi-scale independent features at salient wavelet sub-bands.关键词：melanoma detection;structure irregularity;multi-scale descriptors;salient wavelet sub-bands2719|210|0更新时间：2024-05-08
- 摘要：The problem of dimensionality reduction arises in many fields of information processing, including machine learning, pattern recognition, data mining etc. Locally linear embedding (LLE) is an unsupervised and nonlinear learning algorithm for dimensionality reduction, well-known for its outperformance. Unlike classical LLE, which is based on the L2-norm, a novel L1-norm based LLE (L1-LLE) algorithm is proposed in this article, which is robust to outliers because it utilizes the L1-norm, which is less sensitive to outliers. The proposed L1-norm optimization technique is intuitive, simple, and easy to implement. It is also proven to find a globally minimal solution. The proposed method is applied to several data sets and the performance is compared to those of other conventional methods.关键词：dimensionality reduction;L1-norm;Manifold learning;locally linear embedding;robust3763|253|0更新时间：2024-05-08
- 摘要：For the problem of texture defects segmentation, a new texture image segmentation approach based on the Mean Shift theory combined with the Curvelet transform is proposed. First, a Curvelet transform is used to decompose the image to each channel. Secondly, each channels feature image derived from non\|linear transformation is adaptively clustered to find the singular points using Mean Shift. Finally, the filtered images of all channels are reconstructed to make defects prominent, and the binary image is obtained by a threshold. In this paper, a learning sample is not needed and the multi\|objects boundary is located fast and accurately. This method is robust against rotation, brightness changing, noise, and weak boundaries.The effectiveness of the method is verified by MATLAB simulation experiments.关键词：texture segmentation;curvelet filtering;mean shift2577|187|0更新时间：2024-05-08
- 摘要：Texture is one of the important visual features in image analysis. Gaussian graphical models (GGM) have good prospective, and are applied to construct the texture model. The structure of the GGM is explored by the connection between the local Markov property of texture features and the conditional regression of Gaussian random variables. Thus, the model selection can be converted to select variables. The technique of penalty regularization provides many methods for variable selection and parameter estimation. Furthermore, the methods of penalty regularization conduct neighborhood selection and parameter estimation simultaneously. The texture feature is extracted and applied in texture analysis.关键词：Gaussian graphical models;model selection;texture synthesis;texture classification;texture segmentation3159|220|0更新时间：2024-05-08
- 摘要：Recently, as a difficult problem in video object tracking non-rigid object tracking has received much more attention. To track non-rigid objects exactly and solve the problem caused by shape deformation, this paper proposes a non-rigid object tracking algorithm based on active basis model. Firstly, the object active basis model was learned from the training set using a shared sketch algorithm. Secondly, the learned active model is embedded in a particle filter. The experimental results show that our algorithm is more robust than other methods when non-rigid object has shape deformation and occlusion happens.关键词：non-rigid object tracking;active basis model;particle filter3151|230|0更新时间：2024-05-08
- 摘要：The traditional Mean Shift tracker with single color histogram result in aborting under changes of appearance of object. To deal with this problem, a Mean Shift tracking algorithm using multiple color histograms adaptive integration is proposed. The proposed algorithm enhances the Mean Shift tracker with multiple reference color histograms obtained from different target views, and takes the weighted integration of these histograms as the target model. To adapt to changes of appearance of objects, the proposed algorithm dynamically assesses the reliability of each color histogram and adaptively computes the color’s fusion weight by the ratios of the mean and variance of the probability image of the object. Experimental results show that the proposed Mean Shift tracking algorithm is superior over the existing Mean Shift tracking algorithm of appearance of an object is changing.关键词：object tracking;mean shift;multiple color histograms;kernel function histogram3408|252|0更新时间：2024-05-08
- 摘要：In this paper a new feature point detection method for 3D meshes is proposed. This method serves as an important preprocessing step for a number of 3D applications including mesh simplification, 3D shape matching, and viewpoint selection. Compared with similar algorithms recently proposed, the proposed method has two advantages: 1) our feature point detection algorithm is based on our new perceptual saliency measure, using the local height, rather than being based on traditional curvature. We assume that the perceptual importance of a given point on a 3D model can be described by the protrusiveness of that point, but not the bend degree. Therefore, we proposed the local height as a new measure for evaluating the perceptual importance of a point. 2) We use Mean Shift, a powerful non-parametric estimator of the density gradient, to analyze the distribution of local heights on a mesh, and to detect feature points on this meshExperimental results showed that our proposed method is able to capture perceptually salient feature points on a 3D method,and the algorithm is stable at different levels of details.关键词：3D model;feature point detection;local height;Mean Shift2931|202|0更新时间：2024-05-08
- 摘要：The lack of accordance between the information that one can extract from an image and the interpretability of the same image in a given situation is the most important factor that hampers the accuracy of content-based image retrieval (CBIR). Recently, the combination of several similarity measures draws much interest in the CBIR area, It can be shown that is effective in reducing this discordance.The core problem is:how to choose a better way to combine these similarities? In this paper, we propose a new combination algorithm. It combines similarity measures under the sum rule based on mutual information which estimates the correlation between the continuous random variable similarity measures and the discrete random variable similarity. The experimental results show that this algorithm achieves a high accuracy and efficiency in real-world image collections.关键词：content-based image retrieval;similarity measures combination;mutual information;sum rule2528|249|0更新时间：2024-05-08
- 摘要：Content-based image retrieval (CBIR) is a new information retrieval technology along with development of the digital multimedia technology. The paper proposes relevance feedback techniques and genetic algorithm for image retrieval based on multiple features in allusion to the problem of inaccurate description, low query precision and high Frequency of feedback, which can avoid causing the problem of different images with the same single feature (color, texture and shape). Compared with the existing algorithms, the proposed algorithm can automatically adjust the image feature weights. The experimental results show that the proposed algorithm is robust for rotation, translation and scale changes strongly. In addition, the proposed algorithm has higher query precision and lower frequency of feedback simultaneously.关键词：CBIR;genetic algorithm;relevance feedback;image representation2596|207|0更新时间：2024-05-08
- 摘要：Aiming at overcoming the semantic gap between low-level features and high-level semantic concepts of images and to improve the precision of image classification, we propose an approach to construct a color semantic feature to map low-level features to semantically meaningful categories. Color features are extracted first and then a semantic network containing a set of color concepts and related objects is build. Finally,a color semantic 3-tuple is being constructed.The proposed method can not only get better results in image classification, but is also more robust.关键词：semantic gap;semantic network;image classification;semantic feature2956|245|0更新时间：2024-05-08
- 摘要：We proposes a framework for reconstructing one 3D head model based on two orthogonal images and one 3D generic head model through feature matching. First, we segment the images into face and hair regions. For frontal images, feature points are located by an improved ASM (active shape model), for side images, feature points are found by an improved local maximum curvature tracking method. In this method, we select some distinctive feature points from the generic 3D model. These 3D feature points are matched to the detected image feature points. Based on the radical basis function (RBF), the generic 3D model is morphed, and the shape model is obtained. Since the images from orthogonal views may have different illumination, color transfer is used to enforce the texture color consistency. Based on the shape model, these two images are matched, and one seamless texture is generated. The final head reconstruction is achieved by mapping the texture to the 3D head model.关键词：3D head modeling;feature points matching;color transfer;seamless texture generation;texture mapping3251|210|0更新时间：2024-05-08
- 摘要：Perceptual hashing is a function of mapping from multimedia digital presentations to a perceptual hash value, which provides a secure and reliable technical support in fields such as identification, retrieval, and certification of multimedia content. The current algorithms fail in taking sufficient human visual perceptual factors into consideration. With the improvement of their over-robustness, most of the algorithms can’t assure their securities. In this paper, a novel perceptual hashing algorithm is proposed. In order to simulate multi-channel features of the human visual system, a cortex transformation is combined with a computational model of the human visual system, which is designed by jointly considering four visual perceptual factors during the feature extraction stage, such as spatio-temporal contrast sensitivity function, eye movement, lightness adaptation, and intra-band and inter-band masking. Additionally, a diffusion mechanism is introduced into the preprocessing stage. The results suggest our proposed method could achieve better trade-offs between robust and secure resilient to various content-preserving manipulations, and also reflects the uniformity between subjective perception and objective evaluation.关键词：human visual system (HVS);video perceptual hashing;Cortex transform;diffusion blocking;just noticeable difference3033|207|0更新时间：2024-05-08
- 摘要：A formal language which is simply a set of sequences is composed of some sentences according to certain rules or symbol strings of finite or infinite sets, and formal language theory considers finite descriptions of languages. We are particularly interested in description methods that are easy to understand and use, lead to algorithms for analyzing sequences and are suitable for automated processing. Based on the formal language structure and the grammar rules of regular expressions, an algebraic language systems which can describe L system and iterated function system (IFS) uniformly is proposed by employing algebra. According to the grammar rules generated by the algebraic language systems, Five types of this system production is characterized. Under the algebraic language systems, the DOL system, IFS, condensation IFS, IFSP, and RIFS are mainly described by combining with fractal theory, then fractal attractor is ex?pressed by the regular expression equations of this system, and some examples are given. As examples showed, some fractal graphics can be constructed simply, clearly and efficiently by the language of algebraic system.3846|249|0更新时间：2024-05-08
- 摘要：Straight line is the most basic element of graphics, and it is of great significance to study fast algorithm of line generating. On the basis of Bresenham algorithm, this paper considers and takes full advantage of the relativity between pixels in the same line, and can generate all the pixels in a pixel line every time. The new algorithm based on pixel line can guarantee the line precision to be the same as Bresenham algorithm, but can get higher speed and is easily for hardware implementation.关键词：Straight line;Bresenham algorithm;Pixel line3040|320|0更新时间：2024-05-08
- 摘要：We propose a new approach for detecting edges based on Contourlet probability distribution. First we analyze the mixture of Gaussian distribution traits of the Contourlet coefficients. Then we establish the probability model regarding the Contourlet coefficients which can be described by the big state and the small state. At the same time, we separate the linear singular signals on the model of the image. Afterwards, we improve the maximal between-class variance by a threshold selection method based on the variance between different classes and the in-class variance. This can ensure the maximum distance between classes, while simultaneously increasing the degree of polymerization in a class. Furthermore, we use the threshold for the binarization of the separated singular signals and extract the edge message. Compared to traditional methods, our method detects smooth edges in remote sensing images effectively. Meanwhile, it can shield the unimportant singular information, making it is useful for practical applications.关键词：remote sensing image;Contourlet transform;Gaussian mixture model（GMM）;edge detection2712|224|0更新时间：2024-05-08
- 摘要：This paper proposes a new approach for extraction of main roads from SAR imagery based on parallel pairs detection. The process is divided into three steps, in the low-level step, after the original SAR imagery is filtered to suppress the speckle noise, Ratio Of Exponentially Weighted Averages(ROEWA) operator is used to detect edge; in the mid-level step, eliminating short lines and linking roads firstly, then the approach of parallel pairs detection is carried out; in the high-level step, after linking the roads, we recognize the roads by the proposed model. Finally, the roads are grown according to the conditions. The experiment results show the feasibility of the approach.关键词：SAR image;edge detection;parallel;Ratio of exponentially weighted averages3025|312|0更新时间：2024-05-08
- 摘要：In this paper, we introduce a novel approach to as-rigid-as-possible mesh deformation based on tetrahedral cells. The most distinctive feature of this approach is that users can modify the original mesh by selecting a couple of vertices on the mesh. First, sets of tetrahedral cells are produced, which fit well to the mesh surface, through sparse voxelization. Preserving the rigidity of local transformations of the mesh surface and these tetrahedra during deformation is achieved by minimizing the corresponding energy formulation that prevents unnatural artifacts both on the surface and in the interior of the model. Meanwhile, in order to avoid the possible collapse emerged under large-scale deformation, we present a simple adaptive tetrahedron decomposition method, which brings more local deformation freedom to the interior of the mesh to eliminate the implausible deformation effects based on these dramatic volume changes in the local area. In addition, a few tetrahedra are generated by this method, and then more tetrahedral cells would be densely involved into the deformation process if necessary, there by resulting in highly efficient and robust results. In our experiments, we show that the volume and the surface details of the whole mesh can be approximately preserved and the local degeneration inside the mesh under the large-scale deformation could be efficiently avoided.关键词：shape deformation;rigid deformation;volume constraint;energy minimization;adaptive4805|446|0更新时间：2024-05-08
- 摘要：Owing to the perceptual-saliency measure and saliency-domain shape modeling techniques, a novel shape manipulation scheme is presented to adjust the visually salient features of 3D models. The perceptual-saliency measure for 3D models is introduced, which is defined by a center-surround bilateral filter operator on Gaussian-weighted averages of local projection heights between mesh vertices and their neighbors. The geometric features of the underlying shapes can then be filtered by low-pass filters, high-pass filters, and other enhancements due to the user specified shape manipulation functions defined on the saliency-domain, that is, the low-pass shape manipulation, the high-pass shape manipulation and the enhancement shape manipulation. The experimental results show that it is easy and efficient to achieve different shape modeling results with our proposed technique.关键词：saliency-domain;shape manipulation;low-pass shape manipulation;high-pass shape manipulation;enhancement shape manipulation3997|196|0更新时间：2024-05-08
- 摘要：Bilinear interpolation and various bi-cubic interpolations are frequently adopted in image resizing. However the bi-quadratic function is rarely considered due to its phase distortion problem. In this paper, we propose an image resizing method via bi-quadratic B-spline interpolation, where 4×4 pixels are sampled symmetrically in the local image.\r\nand resamping the image with the function,Thus the phase distortion problem can be avoided. The proposed algorithm is a local algorithm and can be easily extended.According to our experiment results,the proposed bi-quadratic B-spline interpolation algorithm has better image peak signal-to-noise ratio (PSNR) and MISSIM than bi-linear interpolation, bi-cubic convolution, Catmull-Rom cubic interpolation, or the Dodgson interpolation algorithm.The results are comparable to the bi-cubic B-spline interpolation algorithm,though the visual effects are not as good as that,but still better than the Dodgson algorithm. The computing time is reduced by nearly one-third compared to the bi-cubic B-spline interpolation algorithm.Since the algorithm has not carried on the special handling to the image edge features,so the method does not suit the rich details texture images,which we will focus on in the future days.关键词：image resizing;Bi-quadratic B-spline;cubic interpolation6069|219|0更新时间：2024-05-08
- 摘要：When the color of a moving human is similar to the color of the background, one single color space can not very vell describe the differences between the moving human and the background.Faced with this problem,we propose a multi-color space merging strategy. Moving human detection is conducted separately by finding channels which have the best performance, based on a single channel of color space. Moving human detection then can be concluded by merging these channels. Moreover to solve the dynamic background problem, we designed a background updating algorithm which depends on temporal differencing and refreshing. Experiments confirm that in a complex environment, results of our algorithm are better than test results of the algorithm based on a single color space. The algorithm also can deal with dynamic backgrounds effectively.关键词：color space;background updating;temporal differencing;background subtraction4300|234|0更新时间：2024-05-08
- 摘要：Accurate tracking for augmented reality applications is a challenging task. Multi-sensor hybrid tracking generally provides more stable resalts than single visual tracking. A new tightly-coupled hybrid tracking approach combining vision-based systems with an inertial sensor is presented in this paper. Based on the multi-frequency sampling theory in the measurement data synchronization, a strong tracking filter is used to smooth sensor data and estimate the position and orientation. Through adding a time-varying fading factor to adaptively adjust the prediction error covariance of the filter, this method improves the performance of tracking for fast moving targets. Experimental results with occluded markers show that proposed approach can effectively improve the prediction accuracy of location information to target motion with the hybrid tracking algorithm based on the extended Kalman filter, improve the stability of fast moving \r\ntarget tracking. Our approach is suitable for a large range of mobile conditions.关键词：hybrid tracking;multi-sensor;strong tracking filter;augmented reality5042|153|0更新时间：2024-05-08
- 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