最新刊期
卷 18 ， 期 1 ， 2013
- 摘要：With the increasing demand for multi-view face recognition techniques, frontal face image synthesis has been one of the most interesting research topics nowadays. However, it is a classic inverse problem to synthesize frontal images from profiles accurately, and there are some challenges.In this paper,we present a systematical summary of the current frontal image synthesis methods.Furthermore, some classic synthesis strategies are introduced. Moreover,according to the theories used,the methods are classified into two categories:graphic based methods and statistical learning based methods. Additionally, these approaches are compared in three aspects: complicacy, robustness, and performance. At last, some potential pointers towards future research topics are given.关键词：face synthesis;multi-view;inverse problem;review5372|544|0更新时间：2024-05-07
- 摘要：Road network extraction from SAR Images is an important application in the field of remote sensing. With the development of the SAR imaging technology and the urgent demands from remote sensing applications, many road-network extraction methods have been proposed in the recent twenty years.In this paper,we review the development of road extraction methods from SAR images. First, we classify the course of road extraction into two steps: local detection and global connection. Then, some local detection and global connection algorithms are discussed regarding their application fields and their advantages and disadvantages. Finally, the current problems are given and possible future work is explained.关键词：SAR images;road network;local detection;global connection;review3682|328|0更新时间：2024-05-07
- 摘要：Recovering the incomplete fragments of a JPEG file is a challenge in the digital forensics field, especially in the case of where JPEG header is lost or damaged. In this paper,we present a method to decode and display JPEG file fragments whose header has been lost or damaged and its data doesn't include any restart marker. First, the decoding parameters (Huffman tables, image width, quantization tables, sampling factor)are estimated based on the fragment content; second, corresponding adjustment algorithms for the display problems about mismatched positions and colors are proposed. Experiments show that the proposed method can correctly decode and display the JPEG file fragments without header and RST marker.关键词：data recovery;JPEG fragments;restart marker;reconstruction;decoding3768|280|0更新时间：2024-05-07
- 摘要：Fingerprint encryption is one of the most active research domains in biometrics recognition. However, fingerprints are human inherent characteristics, which are associated with one person permanently and cannot be changed. Once one fingerprint template is compromised, it is permanent. In order to avoid the loss of fingerprint templates, a set of irreversible transform functions, which are based on the Butterworth Low Pass Filter, are proposed. They transfer the original fingerprint templates to cancelable fingerprint templates. Those cancelable templates,stored in different recognition system, are different and irreversible, which can overcome the security limitations caused by the invariance and openness of the fingerprint templates. Experiments regarding matching precision, irreversibility and transform efficiency show that this scheme has high performance on authentication, security and computation.关键词：fingerprint encryption;cancelable fingerprint template;Butterworth lowpass filter;irreversible transform function3121|232|0更新时间：2024-05-07
- 摘要：An image super-resolution (SR) reconstruction algorithm based on residual error is proposed. Patch pairs, composed of features for low-resolution (LR) patches and residual errors between original high-resolution (HR) image patches and interpolated LR image patches, are classified by K-means, Each class patch pair is trained by KSVD (K-singular value decomposition) to obtain an LR and HR dictionary pair. Residual errors are reconstructed by the dictionary pairs selected by the Euclidean distance between the test patches and class centers and by the weighted sum of the reconstructed results of the similar class patches. Then, combined with interpolated LR images and reconstructed residual errors, HR images are reconstructed. Experimental results show that the proposed method has a better performance and the method to classify patches and perform weight sum of the reconstructed results of the similar class patches is improving the quality of the SR image.关键词：residual error;super-resolution;dictionary;K-singular value decomposition3651|511|0更新时间：2024-05-07
- 摘要：In this paper,we propose an effective algorithm for texture synthesis using rotational Wang Tiles, which can solve several problems of Wang Tiles,especially the incomplete sampling, the non-optimal cutting path, and the error of the center and corner areas. The initial framework of a rotational Wang Tile is constructed using four square sub-images. The overlapping region of the texture is divided into two types and the shortest path algorithm based on edge data structure finds the cutting paths. The large texture is created by stochastically filling rotational Wang Tiles. The experimental results show that the algorithm for a variety of textures can be obtained high quality synthesis results in real-time.关键词：texture synthesis;Wang Tiles;cutting path;tiles rotation3449|245|0更新时间：2024-05-07
- 摘要：The microscopic image of protoplasm somatic cells typically has blurred boundaries and inhomogeneous object regions.Therefore, it is difficult to segment the cells using traditional methods. First, because the protoplasm cell is round, the circular prior knowledge is added to the fast level set method and then a new circle dependent fast level set segmentation method is proposed. Then, to solve the problem of segmentation for multi protoplasm somatic cells, the pre-segmentation is used before using the fast multi-level set method based on circle information. Furthermore, a new fast level-set method, which is based on the histogram, is proposed to get a better result for the pre-segmentation. The eight-chain code tracking method and the randomized Hough transform for circle detection are used to divide object region resulting from pre-segmentation respectively for multi-cells and multi-clustered cells. Finally, experimental results show that the new method proposed in this paper can deal well with the problem of segmentation for protoplasm somatic cells.关键词：fast level-set method;protoplasm;circular prior knowledge;multi-circle segmentation2843|410|0更新时间：2024-05-07
- 摘要：Video fingerprinting techniques have many applications in video retrieval, identification, and security. A robust video fingerprinting based on compressed sensing is proposed. In video fingerprinting, video key clips extracted are sampled by using the sparse and safety of compressed sensing. Each matrix sample is made into blocks, from which several big energy blocks are made into a new feature matrix. The singular value is used as code for the fingerprinting by SVD from this new feature matrix. Furthermore, an efficient two step match algorithm is proposed to using a search and match approximation, which improves video searching speed. The experimental results show that the proposed video fingerprinting is accurate in identifying different video clips, robust against common video processing, and can retrieve videos in real-time.关键词：compressed sensing;video fingerprinting;video identification;singular value decomposition2406|240|0更新时间：2024-05-07
- 摘要：An effective approach for visual saliency detection can help people search for the object of interest from vast visual information rapidly and accurately. Considering the complexity of noises covering a wide area in actual road images, we present a new pavement crack detection approach based on image saliency in this paper. This approach calculates the salient value of crack images in a coarse scale based on the grayscale sparsity and global contrast after grayscale correction on images, which are divided into small blocks. Then, according to the characteristics of the cracks, such as local brightness, edge, and continuity, we calculate the local saliency in the continuously outspread local neighbor domain in a fine scale. After enhancing the saliency based on the spatial continuity, we extract cracks using adaptive image segmentation method. A large number of experimental results demonstrate this approach can detect the crack areas more correctly and effectively compared with traditional methods. It better suppress noises, has lower missing rate and misuse detection rate. Moreover, the result is consistent with human visual characteristics.关键词：crack detection;image saliency;saliency enhancement;grayscale correction4838|701|0更新时间：2024-05-07
- 摘要：Using the C-V model to segment images with intensity inhomogeneity, the segmentation results are often not very good. Therefore, we propose an active contour model based on the local entropy energy. First, we introduce the concept of local entropy into the C-V model to get inhomogeneity information in local regions according to the kernel function and to model the local entropy energy function. Second, we use a variable level set to minimize the local entropy function and to get the gradient descent flow of the level set. Finally, simulation experiments are carried out on four severe intensity inhomogeneity images, and the results are compared to the proposed method with LBF and LGDF methods. It is shown that our method achieves more accurate segmentation results for intensity inhomogeneity images compared to the LBF and LGDF methods.关键词：C-V model;image segmentation;intensity inhomogeneity;entropy;level set;gradient descent flow3043|302|0更新时间：2024-05-07
- 摘要：In recent years there has been a great interest in the study of morphological decomposition of image into cartoon and texture, and its hierarchical decomposition. The existing algorithms suffer from high computational complexity and are not very well suited for decomposing images hierarchically. We propose an operator-based image decomposition method, in which a local linear singular operator characterizes the texture component and the residual signal is treated as a cartoon component. With proper changes to the adaptive parameter, it can decompose the image hierarchically through the decomposition of the residual signal. The adaptive parameter is determined according to the change rate of local total variation. The experimental results demonstrate that this method is effective and shows better visual effects.关键词：integral operator;differential operator;image decomposition;cartoon texture;hierarchical decomposition2678|286|0更新时间：2024-05-07
- 摘要：To satisfy the stringent requirements of the object tracking performance in the robot's learning-from-demonstration-framework, a new tracking algorithm that can deal with fast motions, occlusions, and drifts, is proposed. First, the Median-Flow method is used to predict the position-shift of the object and the Gaussian weight of each patch. Then, the search-region is modified and the object is located by the online multi-instance learning classifier. Afterwards, the likelihood of each patch is calculated. Finally, the results are combined under the Bayes framework to get the best prediction by exhaustive search and the online classifier is updated. Experiments in several commonly used test videos show that our method outperforms the other state-of-the-art tracking methods, especially for fast motion and drifts. Furthermore, the proposed method can run in real-time.关键词：service robots;learning from demonstration;object tracking;online multi-instance learning;median flow3415|296|0更新时间：2024-05-07
- 摘要：The formal model of the spatial directional relation is one of the most important parts in the area of spatial relation. However, the research on the expression method of special directional relation is still immature. In this paper, the characteristics and defects of the recent special directional relation models are analyzed first; Second the basic idea, and the construction method of the quadtree histogram are introduced; third a new expression method to judge the special directional relation based on quadtree histogram is described in detail; finally, to prove the correctness of the method, several contrast experiments are given based on synthetic and real images. The experimental results show that the model can get close to the verdict like human cognition for special directional relation and overcome the problems of the existing models.关键词：spatial directional relation;special directional relation;quadtree histogram;fuzzy sets2714|212|0更新时间：2024-05-07
- 摘要：Texture evolutionary system is a near-regular texture oriented synthesis algorithm. The main feature of the system is the definition of behaviors, which, based on the evolution theory, minimize the error accumulation problem caused by the inflexibility of synthesized blocks.In this paper, we present a new texture synthesis algorithm on the concept of co- evolution. With the new way of individual selection and arrangement, it can be applied to the near-regular textures in any period direction. The approach of removing the settling behavior and pre-computation reduces the compute redundancy. In addition, the new process of evolution can be better parallelized. Results show that the proposed algorithm does not only reinforce the versatility but also improve the efficiency greatly.关键词：texture synthesis;near-regular texture;evolutionary system;co-evolution2861|327|0更新时间：2024-05-07
- 摘要：Gray-conversion is a time-consuming operation for drawing anti-aliased straight lines or curves. In order to improve the efficiency of drawing anti-aliased straight lines, we propose a new anti-aliasing algorithm based on a two-pixel model. Different from other anti-aliasing algorithms, a technique which controls iterations by grey values directly rather than distances is proposed. Calculations about distances between pixels and real line and distance-grey conversion are abandoned, but accurate grey values rather than estimated ones can be obtained. The analyses prove that for one step only 4～5 integer operations are invoked in this algorithm, and the drawing speed is faster than other methods. Furthermore, this algorithm can be constructed with the simplicity similar to basic scan conversion algorithms.关键词：straight line drawing;anti-aliased;scan-conversion;integer operations;grey-iteration control;grey-conversion2949|229|0更新时间：2024-05-07
- 摘要：Micro-blog messages usually contain a great deal of traffic information such as traffic conditions, traffic events and traffic controls, which can be useed as a complement to conventional traffic information collection technologies like fixed sensors and floating cars. However, due to ambiguous narrating, uncertainty, and the unstructured characteristics of micro-blog messages, extracting traffic information from micro-blog messages is rather difficult. In this paper, we propose an approach for extracting traffic information from a large amount of micro-blog messages. First, we build a traffic information table by semantically extracting traffic related words from micro-blog messages and matching each word onto the corresponding road segment of the road networks. Then, according to the traffic information table, we evaluate the highest confidence level of traffic condition for each road segment by using a neural network based Fuzzy-C-Means (FCM) clustering method, to obtain the most confident road conditions. Experiments on Beijing road networks with a large number of Sina micro-blog messages verify the effectiveness of the presented approach.关键词：micro-blog;traffic information;word segmentation;fuzzy clustering;clear degree;degree of confidence3773|241|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