最新刊期
卷 17 ， 期 11 ， 2012
- 摘要：Digital image compositing is receiving increasing attention from the image processing research community, and it has been used in many applications such as photo editing, graphic design, and for visual effects in movies. The essential of image compositing is how to extract foreground object from a given image and composite it to a new background seamlessly. In this paper, we give an overview over the different approaches of digital image compositing, which are categorized as matting-based image compositing, gradient-based image compositing, and multi-resolution-based image compositing. The representative methods in each category are stated briefly, and then we compare and analyze their performance according to quality of compositing, robustness and computation time. Next, several state-of-art applications of image compositing are introduced as extension of this survey. Finally, the limitations, the research challenges and the future directions of image compositing are discussed.关键词：image compositing;matting;Poisson cloning;seamless compositing4656|175|0更新时间：2024-05-08
- 摘要：Sparse representation of signals is an evolving field in many machine learning and image processing tasks. Nowadays, more and more attention is paid on the algorithm for learning dictionaries.Traditionally, the dictionary is an unstructured set of atoms. Considering the sparsity of the group of the sparse representation signal, a mathematical model of the dictionary learning based on the group sparsity is constructed. We propose an efficient algorithm for learning structured dictionary according to the convex analysis and monotone operator theory. The experiments show that the algorithm converges faster, the dictionary trained from the new model adapts better to the data and the data is better represented, which overall improves the image enhancement effect.关键词：sparse representation;dictionary learning;group sparsity;convex optimization;monotone operator4859|100|0更新时间：2024-05-08
- 摘要：The performance of depth map coding in 3D videos mainly depends on the subjective quality of the rendering view. Traditional quality metric, the mean square error(MSE), doesnot characterize the human visual system (HVS)well. In this paper, we present a depth map coding method based on virtual view rendering distortion using structural similarity index (SSIM)as a quality metric. By analyzing the relationship between distortions in the coded depth map and distortions using SSIM as quality metric in the rendered view, a distortion model that estimates the rendering distortions caused by depth changes on depth coding is proposed. The rendering distortion estimation model is applied to the perceptual rate-distortion optimization (RDO)framework based on SSIM which decides the best mode in depth map coding. Experimental results with the proposed method show higher SSIM and better subjective quality for the rendered views, while reducing objects boundary’s distortion and artifacts in rendered view.关键词：depth coding;view rendering;rate-distortion optimization;distortion analysis2949|42|0更新时间：2024-05-08
- 摘要：Robust lossless data hiding has extensive applications in medical imaging systems, law enforcement, and remote sensing. A robust lossless image data hiding scheme is proposed in this paper. The original cover image can be recovered without any distortion after the hidden data has been extracted if the stego-image remains intact, and conversely, the hidden data can still be extracted correctly if the stego-image goes through JPEG 2000 compression to some extent. First,the proposed scheme divides a cover image into a number of non-overlapping blocks and calculates the statistical quantity of each block. Then the statistical quantity values are shifted by appropriate thresholds, which are selected according to the maximum absolute value of the statistical quantity. Finally, we can embed secret bits into the blocks by using the statistical quantity values which have been shifted. Simulation results demonstrate that the proposed algorithm can achieve high performances in the visual quality of the stego-images, data embedding capacity, and robustness. Performance comparisons with other existing robust lossless data hiding schemes are provided to demonstrate the superiority of the proposed scheme in embedding capacity when keeping the visual quality and robustness at a similar level.关键词：robust lossless data hiding;statistical quantity;visual quality of images;embedding capacity;robustness;JPEG2000 compression2922|53|0更新时间：2024-05-08
- 摘要：In image forging, geometric transformations, JPEG compression, and blurring are typical operations. In this manuscript, algorithms for detection of the typical operations in a forged image are proposed based on operational characteristics. First, a possibly composite image is divided into overlapping blocks, and a block measure factor is defined and adopted to describe both re-sampling and JPEG compression characteristics for each block, followed by detection of tampered regions. Experimental results show that compared with the existing single targeted detection methods, the proposed algorithm can recognize forged images under more combinations of tampering modes and the tampered regions are located more effectively. Furthermore, the proposed method performs well even when the JPEG quality factor is small. Second, an approach is proposed to detect blurring traces. The image is blurred again with an appropriate blurring kernel and the difference of image pixels are estimated before and after double blurring. The tampered regions may be detected through the classification of the values in the difference image. Simulation results show that the algorithm is effective for blurring detection with various blurring operations and it also robust against lossy JPEG compression. Comparing with the existing block-based methods, the proposed method can reduce the computational complexity greatly by avoiding the point-by-point block calculation, and can detect small fuzzy traces.关键词：forged image;re-sampling;JPEG compression;blur operation;tampering detection3188|114|0更新时间：2024-05-08
- 摘要：Because of the low contrast, lack of color, and low dynamic range of infrared images, the object tracking based on infrared imaging is rather difficult.An infrared object tracking algorithm is proposed by integrating the gray kernel histogram and SURF (speeded up robust features)features. An object template is represented by gray kernel histogram and SURF features in the first frame. The Mean Shift algorithm is used to find the suboptimal position rapidly in the next frame. Because the gray histogram contains less information, the tracking error is accumulated. Then, the improved SURF feature matching algorithm is used to estimate the size and center point of the current frame. The cumulative errors are amended to avoid the tracking window drifting gradually away from the object and the size of tracking window can be self-adapted. Finally, the object template is updated. Experimental results on real situations demonstrate that the proposed algorithm can track objects well in real-time ever when the appearance changes and similar apparents are existing around the targets.关键词：infrared object tracking;kernel histogram;speeded-up robust features(SURF);mean shift2819|85|0更新时间：2024-05-08
- 摘要：A front extraction method for a raster image of sea surface temperature (SST)is presented. Considering that the front was distributed unevenly, a low-pass filter is used to smooth the SST gradient image. The gradient image is then divided into target and background parts by a split threshold calculated from the iteration algorithm. By mathematical morphology image thinning, a SST front skeleton is extracted from the target one, and the tiny branches are trimmed. After the vectorization of the SST front skeleton, the vectorized SST front lines are smoothed using the erasing angle method. An example of SST front extracting process in the west Pacific is presented at last. The results show that the method is feasible and effective.关键词：sea surface temperature;mathematical morphology;front extraction;vectorization3153|244|0更新时间：2024-05-08
- 摘要：In this paper, we present a real-time background modeling framework for moving vehicle segmentation in traffic surveillance scenes. First, we propose a new adaptive background modeling method called Balloon Model established in the color space, which fixes the error of non-moving objects in the parameter models; second, an effective shadow detection approach is developed to detect shadows in traffic scenes, which adopts feature fusion method to achieve the segmentation of vehicles. Compared to the other methods, this technique achieves a higher accuracy and is faster.关键词：balloon model;adaptive background modeling;moving object detection;shadow detection3022|124|0更新时间：2024-05-08
- 摘要：In order to overcome the semantic gap between low-level features and high-level semantic concepts of imagery, a new image annotation refinement approach based on Random Dot Product Graph (RDPG)is proposed. In our approach, the visual features of images are used to construct a semantic graph of the candidate annotations. Then, we reconstruct the semantic graph with a RDPG, find the unobserved relevance in the incompletely observed semantic graph, and transform the random graph into the probabilities of state transition. Combined with Random Walk with Restart (RWR), the final annotations are chosen. This new method incorporates the visual and semantic information of images, and reduces the influence of the scale of database. Experiments conducted on three standard databases demonstrate that our approach outperforms the existing image annotation refinement techniques. The macro F-Score and micro average F-Score can reach 0.784 and 0.743 respectively.关键词：image annotation refinement;random dot product graph(RDPG);semantic graph;random walk with restart(RWR)2689|38|0更新时间：2024-05-08
- 摘要：In this paper,we present a new scene categorization algorithm based on supervised subspace modeling and sparse representation. The proposed method implements supervised dictionary learning via decomposing the unsupervised sparse dictionary learning model into a group of independent optimization problems. After learning the dictionaries of all categories, we aggregate them to form a global dictionary and encode each local feature of an image based on it. After using spatial pyramid representation and max pooling of local features’ coding vectors, the final holistic feature depicting a scene image can be retrived. Comprehensive experimental results on four popular benchmark scene datasets show that our method achieves very promising result compared to existing state-of-the-art techniques.关键词：sparse representation;dictionary learning;spatial pyramid matching;scene categorization;subspace modeling3060|75|0更新时间：2024-05-08
- 摘要：The target representation method of a tracked target has great influence on the robustness of the tracking algorithm. In this paper,we introduce a new texture feature called Opponent Color Local Binary Patterns (OCLBP). By analyzing the correlation among different color channels and all the ten texture patterns of the OCLBP, we select the texture histogram of the key points which correspond to only the seven major patterns of the OCLBP to represent the target candidate region.Finally, this model is integrated into the mean shift framework for object tracking. The experimental results illustrate that the proposed major OCLBP patterns based method can significantly improve the performance of Mean Shift object tracking algorithm.关键词：object tracking;mean shift;OCLBP;key points2806|74|0更新时间：2024-05-08
- 摘要：Shape decomposition usually plays a significant role in shape analysis and its various applications. In this paper, we present a shape decomposition algorithm that combines the strength of skeleton and boundary features, which carry both global and local information of object shapes. In the proposed method, a bending potential ratio is introduced as a constraint to generate controllable decomposition results. Besides, the algorithm is able to avoid conglutination of important parts by fully utilizing the discrete curve evolution information on the boundary. Furthermore, the adopted robust skeleton method ensures noise insensitive decomposition results and avoids decomposing important parts into trivial ones. We choose the MPEG7 shape dataset and other traditional testing shapes as our experiment data. Experimental results show that our method satisfies subjective visual perception on shape decomposition and is robust to large shape noise.关键词：shape decomposition;skeleton extraction;discrete curve evolution;bending potential ratio2724|51|0更新时间：2024-05-08
- 摘要：In order to render large-scale terrain on graphics processing units (GPU)using hardware-accelerated tessellation, a screen-space adaptive tessellation algorithm for terrain rendering is presented. The triangulation is performed entirely on the GPU, based on analyzing the principle of terrain tessellation. The proposed approach organizes the terrain data hierarchically by tiles and patches, which is the base for a terrain LOD simplification approach processed separately on the CPU and GPU. The edge-based tessellation LOD model for each patch is constructed to compute the tessellation factors in the Hull Shader for the water tightness surface. The procedure for terrain displacement mapping in the Domain Shader is designed to offset and transform each vertex height. Furthermore, a two-level view frustum culling mechanism is used to minimize the data to be rendered. The experimental results show that the algorithm has better screen-space adaptivity and rendering performance. It can produce the terrain model with high resolution geometric details in spite of inputting coarse triangle meshes.关键词：terrain rendering;tessellation;displacement mapping;level of detail;graphics processing unit3522|48|0更新时间：2024-05-08
- 摘要：In this paper,a new method for filtering SAR images using sparse optimization model is proposed. The algorithm based on sparse representation via an over-complete dictionary has a strong data sparseness and provides solid modeling assumptions for the data sets. First,a sparse optimization model based on structural properties of then SAR image is built by regulation. Second,a practical optimization strategy is used to design a redundancy dictionary. Then,an over-complete dictionary is constructed by employing a combined dictionary consisting of wavelets,shearlets,and a redundancy dictionary. Finally,the filtering process is realized through the solution of the multi-objective optimization problem in which the mean backscatter power is reconstructed. The experimental results demonstrate that the proposed algorithm has good de-speckling capability and better enhances image details.关键词：filtering;SAR image;sparse optimization;wavelets;shearlets3015|61|0更新时间：2024-05-08
- 摘要：Due to sensor-to-sensor variation within instruments,stripe noise,which affects image quality and subsequent quantitative calculation,is often detected in remote sensing data. Most previous destriping methods are based on the assumption that photomulipliers are linear. In fact,the nonlinearity is stronger in the low and high signal regions. Moment matching is emphasized in detail and a piece-wise linear dynamic moment matching algorithm is suggested which thresholds the image into low-median-high regions,and destripes each subscene separately by dynamically using its neighborhood average value and standard deviation as reference values. This is equivalent to modeling the relationship between sensors as piece-wise linear rather than simple linear. Tests on band 4 of a TM image and on HJ-1A HSI data show that piece-wise linear dynamic moment matching algorithm reduces stripes to a greater degree while retaining the basic information of image than dynamic moment matching method,Fourier transformation method and automatic equalization curves method. The visual and quantitative assessments make sure that this method is reliable and improves destriping effect of huge water body in heterogeneous area.关键词：destriping;moment matching;piece-wise linear;TM data3050|30|0更新时间：2024-05-08
- 摘要：Image matching is a fundamental step in remote sensing image processing and analysis. The traditional gray correlation coefficient matching algorithm does not have the rotation invariant feature. SIFT (scale invariant feature transform)algorithm can provide robust matching which is invariant to image scale and rotation. However, for high-resolution remote sensing images with clearer geometric structure and richer texture information, the problem of consuming large memory and slow computing is very prominent. In this paper, the image matching algorithm based on Harris corner and SIFT descriptor is proposed. The experimental results show that, compared to the SIFT algorithm, this algorithm greatly reduces the running time. It preserves the invariance of rotation and change in illumination by using SIFT descriptor, overcomes the shortcomings of the gray correlation coefficient matching algorithm, and has good performance on high-resolution remote sensing image matching.关键词：SIFT descriptor;Harris corner;high-resolution;image matching3319|90|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