最新刊期
卷 13 ， 期 10 ， 2008
- 摘要：To detect the tiny sclerotic tissues in pathological examination, the α stable distribution and the fractional lower order statistics (FLOS) (i.e. fractional lower order moment, FLOM) are introduced. The threshold detection method based on FLOS under the non Gaussian condition, the definition of the granularity distribution function and the autocorrelation moment are proposed as well. Firstly, taking the supersonic medical image as an example, the method based on Gaussian distribution and the α stable distribution method respectively to carry on the threshold detection for the supersonic medical image are discussed. After the detection, mathematics morphology is applied to calculate granularity distribution function of the horizontal direction and the vertical direction respectively. At last, the correlation coefficients of two directions are obtained. The experimental results show the threshold detection and granularity analysis method based on the fractional lower order statistics has its superiority.3832|296|0更新时间：2024-05-08
- 摘要：This paper presents a pedestrian detection approach using spatial histograms of oriented gradient feature. In order to detect pedestrians faster and more accurately, histogram similarity and Fisher criterion are employed to measure discriminability of all features and then we selected some discriminative features to identify the pedestrian. SVM classifier is constructed to train the selected features from the target and surrounding background. Experimental results show that the proposed approach is efficient and rapid in pedestrian detection.3778|270|0更新时间：2024-05-08
- 摘要：Digital film archives are often damaged by aging and multiple playing, such as dust spots, dirt, blotches, film unsteadiness, line scratches, flicker, noises, color variations and blurs, witch may cause some artifacts on the films. This paper introduces a method for blotch detection. The initial detection is gained by using an improved SROD algorithm. And then postprocessing module is achieved through two EM steps, thus eliminating the effect of noise and determining blotch. Experimental results show that the blotch detection method proposed in this paper is effective and can be easily implemented.3256|231|0更新时间：2024-05-08
- 摘要：JPEG2000 compression algorithm has broad applications in the field of satellite remote image compression. However, the lifting coefficients of CDF9/7 are a group of irrational numbers, which result in complex computation and require vast storage directly used in lifting algorithm, plus a fixed point method is needed. The existed fixed point implementation methods of CDF9/7 can solve the problem of computation and storage, but they degrade the quality of the reconstructed image. In order to solve this problem, a fixed point implementation method of LS9/7, based on constructing LS9/7 wavelet bases and adopting post scaling method, is proposed in this paper. Compared with two fixed point implementation methods of CDF9/7, the complexity of computation has been decreased by 34% and 28.3% and the quality of reconstructed image has been improved by more than 3db. Its performance is very close to the float point implementation of CDF9/7.3075|200|0更新时间：2024-05-08
- 摘要：We propose a FCM based image segmentation algorithm using multi color components. Firstly, the image is converted from RGB color space to HSI color space, and initial clustering centers of H component and I component are selected according to the data distribution. Then, the FCM algorithm is performed on the H component and I component, and we can obtain the image pixel membership for H component and I component. Finally, two dimensional image features are constructed with the image pixel membership, and the FCM based image segmentation is performed using two dimensional image features. Experimental results show that the proposed method is simple and work well for most images, and has better segmentation effect than the existing FCM color image segmentation.3415|302|0更新时间：2024-05-08
- 摘要：Considering the dependencies between the coefficients and their parents, a non Gaussian bivariate distribution model is given in non subsampled Contourlet transform domain. A novel non subsampled Contourlet transform segmentation method based on the bivariate model is proposed. In experiments, synthetic mosaic image and real images were selected to evaluate the performance of the method, and the segmentation results were compared with wavelet domain hidden Markov tree model method and contourlet domain hidden markov tree model segmentation method. The simulation results indicate that the proposed method has better performance, such as keeps better visual result and reserves more information in edges. As a simple model, the time complexity for model training is lower than other models in comparison experiments.3620|233|0更新时间：2024-05-08
- 摘要：Kernel based Support Vector Machine （SVM）does not consider inner property of training data, so classification results are usually not in optimum condition. In this paper we present a new SVM classification algorithm. The proposed method alters the kernel based on the class information of the training data, with input vectors being classified by this transformed kernel. The described algorithm can improve performance of mapping function indirectly. Simulation and experiments validate that it can improve classification performance and robustness, and reduce noise.3622|203|0更新时间：2024-05-08
- 摘要：The paper proposes the adaptive search range algorithm of UMHexagonS, which is one of motion estimation algorithms for video encoding system on the H.264 standard. Based on analyzing the search range theoretically and experimentally, its search range algorithm can be improved. Then, the adaptive algorithm, derived from the anterior search points, is created to reduce the redundant search points. The result shows that the adaptive algorithm can reduce the 15% period of motion estimation under the premise of little change on PSNR and Bit Rate.3340|218|0更新时间：2024-05-08
- 摘要：A generalization of Linear Active Discriminant Functions named as Nonlinear Active Discriminant Functions (nonlinear ADF) to deal with nonlinear deformations of handwritten character is proposed. In Nonlinear ADF, Kernel PCA is applied to capture and represent the nonlinear deformations. Input space is mapped to feature space through nonlinear mapping. Then an optimal active prototype model is produced in principal subspace of the feature space and the distance between it and the projection of character feature vector in the principal subspace is defined as Nonlinear ADF. In addition, the Nonlinear ADF is further optimized using Minimum Classification Error criterion. Experimental results demonstrated that Nonlinear ADF has achieved a higher recognition rate than that of Linear ADF.3291|213|0更新时间：2024-05-08
- 摘要：Finger print recognition technology is practical and feasible. It is a valuable ID recognition technology. The algorithm we commonly use now cannot provide reliability and high efficient at the same time. Based on effective area capturing and ridges tracing, this article advances an efficient fingerprint extraction algorithm. To capture the effective area and lower down the dependence to the image quality,we first check the core point. Then the tracing ridge is used to extract strange points. Experimental results show that this algorithm can balance the practicality and feasibility at the same time.3257|263|0更新时间：2024-05-08
- 摘要：Pedestrian detection is a difficult problem in the field of object detection. We combine Kobi Levi and Yair Weiss’s edge orientation histogram and Dalal’s hog(histogram of gradients) feature and apply them to pedestrian detection. We improve the algorithm from the following aspects first, we have changed the calculation formula of the original EOH (edge orientation histogram) to gain more descriptive ability. Second, we have changed the policy of updating the weight of the samples of the original Adaboost algorithm in order to reduce overfitting. Experiments show our method is very efficient. When the false positive rate is 1/10000, our detection rate is about 90% on Inria pedestrian dataset. The running speed is about 2 fps with 640×480 images on a 1.8GHz CPU.3864|233|0更新时间：2024-05-08
- 摘要：Automatic age estimation based on facial images is important but challenging in face recognition research. An improved NMF (Non negative Matrix Factorization) algorithm was proposed to implement the age estimation of facial images, which employs linear discriminative analysis in the features derived from NMF, keeps down the base images that have the best discriminate ability to form a new subspace. Then, after projecting the whole training sets images to the obtained subspace, the RBF neural networks was used to extract the aging information contained in most facial images to estimate age information. Finally, experimental results demonstrate that it is an effective method.3647|239|0更新时间：2024-05-08
- 摘要：The author puts forward one object tracking method based on block motion estimation which can overcome sudden light change and obtain good tracking performance in dynamic background. Firstly, block motion compensation and light compensation are processed according to the parameters acquired by block Harris corner match. Then based on the multi feature fusione, combining the color histogram and HOG as the object feature, the algorithm will search for the candidate, with the math point calculated with inter frame difference as the original searching point, which matches the model in the best way. The experiments show that this algorithm can realize accurate and real time tracking in dynamic background and can overcome sudden light change as well.3005|229|0更新时间：2024-05-08
- 摘要：To track multiple objects which were occluded each other, a multi target tracking method based on the sampled corners was proposed, which extracted Harris corner information as the training sample before occluding and used K nearest neighborhood (KNN) classifier to classify the corners in the occluding region. And this method can effectively distinct multiple targets through occlusion. In the process of extracting corners, the dual threshold Harris corner detection algorithm was proposed. It can adaptively and accurately extract corners. The experimental results show that the proposed method can effectively track multiple objects which were occluded and didn’t split up.3513|241|0更新时间：2024-05-08
- 摘要：A new small target detection method for infrared image based on energy accumulation and mean shift clustering is presented. Firstly, accumulating the energy of the infrared image sequence in the sliding window with fixed size can remove the random noise in the infrared images to increase the SNR (signal to noise ratio). Secondly, we suppress the background by using the difference of adaptive mean shift filtering results with two different bandwidths. And thirdly,we segmenting the regions where there are possible targets and extract the statistical features. Finally filter the true targets according to the principle of moving continuity and trajectory consistency of moving target in the image sequences. Target image and background image can be obtained by use of differernt mean shift clustering, the noise can be excluded at the same time. The experimental results indicate that the method can effectively detect the small and weak targets under complex backgrounds.3265|275|0更新时间：2024-05-08
- 摘要：Choosing a distinctive feature and matching criterion is key to developing a reliable face recognition system. This paper discusses the availability of one of geometric feature invariants, scale invariant feature transform (SIFT) descriptor based face recognition. The SIFT feature description of an image is typically complex. In most cases, the difficulty of feature matching problem is aggravated when the different face expressions and image blur exist. For abovementioned issues, in this paper we proposes a new method that six interest sub regions from the face are selected to be described and later be calculated through different weights according to their distinctiveness. The square of the similarity is used to solve the problem of data deviation. The experimental results demonstrate that our method does effectively moderate the face expression effect. It also successfully reduces the complexity and matching time of SIFT feature sets.4111|300|0更新时间：2024-05-08
- 摘要：Visual attention is one of the most important mechanisms of the human visual system (HVS). Recent research has demonstrated that a bottom up visual selective model can be applied to problems such as target recognition. Nevertheless, an image can not be fully described only through a visual selective model because a salient feature can become less salient in certain situations. Humans may become attracted by features which are in minority. This paper proposes a way of combining visual selective model with global rarity to group together images. Experimental results show that the proposed approach works well for image classification and the average accuracy rate can reach 9774%.3430|271|0更新时间：2024-05-08
- 摘要：In this paper, a novel approach to semantic concept detection based on local color spatial feature is proposed. There is much noise and redundant information in many global features of color, texture and shape. This local color spatial feature contains more semantic contents of image than other global features by using prior knowledge of semantic concept level to reduce feature dimensions. Experiment results are reported and presented to demonstrate the effectiveness and efficiency of the proposed approach. Average precision of image retrieval by using the semantic concept detection method based on local color spatial feature is 364% higher than the method based on global color feature.2946|243|0更新时间：2024-05-08
- 摘要：The image transition region is a particular area that ubiquitously exists in image. Because of the existence of the image transition region, causing low accuracy in image segmentation and misinterpretation of image itself. This paper used the method of the local complexity which could accurately draw the transition region, and confirm the processing direction of the image transition region. The experimental result shows that the impact from the image transition region is reduced while edges are much clearer after processing. This method can keep the original internal details of the image and has higher practicability and popularization.3281|219|0更新时间：2024-05-08
- 摘要：To extract the contour features of pedestrian under surveillance, this paper applies Minimum Description Length Shape to get compact point correspondences, which can capture important shape characteristics, of 2D+time contour, and then uses Elliptic Fourier Decomposition to a series of EFT coefficients at different frequency level to compose a contour vector. Experiments indicate our method is efficient in contour representation and vector dimension reduction.3005|239|0更新时间：2024-05-08
- 摘要：In order to automatically detect garment manufacture defects, this paper researches on knitted fabric stitching manufacture inspection. A method based on Genetic Algorithms (GA) to search the optimal structuring element is presented. Morphological filtering technology processing with the optimal structuring element is used to prevent the normal fabric structure from disturbing image segmentation. A binary image representing the property of seams or manufacture defects is obtained using gray thresholding segmentation. According to the experimental results, the method is effective.3067|250|0更新时间：2024-05-08
- 摘要：Moving detection is a key technology in robust video surveillance. Currently widely used Gaussian mixture model(GMM) always detects incorrectly and cannot deal with shadows based on the pixel level and time domain classification, so we introduce an effective algorithm extending the pixel level detection to space domain detection with the combination of illumination of the pixel using GMM and apply it for shadow removal after the first step when foreground pixels has been got. Besides, some parameters in the standard GMM are modified. Experiments show that our algorithm is effective both on detect accuracy and shadow removal.3471|249|0更新时间：2024-05-08
- 摘要：Based on ENVISAT ASAR data in Changle test area, this paper studies the method of rice information extraction. With the analysis of rice backscatter coefficient change rule based on two dual polarized ASAR images, a new band image was constructed to enhance rice information and reduce confusion between rice and non rice classes. Supervised classification was carried out here to extract rice information, and the extraction result achieves a 94.92% overall accuracy. The study indicates that the proposed method can be used effectively for rice extraction.2861|163|0更新时间：2024-05-08
- 摘要：In this paper we propose three types of features which can be used to realize efficient universal steganalysis on color images. After measuring the deviation of DCT coefficient distribution, the correlation of smoothness in spatial domain, and the correlation between different components, a 10 dimentional feature vector is produced for each color image. Those features are sensitive to steganographic process and the feature vector, having a low dimension, is easy to calculate. Then, a SVM classifier is built to distinguish the stego images from the original images. Experiments show that the universal steganalytic approach can effectively detect the presence of secret message hidden by Jsteg, F5 or MB steganography, and can identify the steganographic techniques used.3097|226|0更新时间：2024-05-08
- 摘要：The main focus in this paper is the detection techniques based on the cover images, which are detected using FCM OC SVM. The feature set of cover samples are firstly clustered by the FCM algorithm. Then, the sub class data are trained separately and the multi hyper spheres classification models are established. This technology can improve the detection of cover image and stego image and decrease false detection. Meanwhile the effect of many coefficients on the detecting accuracy is analyzed and generalized for broad application.3035|216|0更新时间：2024-05-08
- 摘要：Camera calibration is a key technology in computer vision, and multi camera calibration is the inevitable step towards an efficient use of multi camera systems. For multi camera systems, a method based on multi view registration is proposed. A planar model with known reference points placed at a few different locations is only required as a calibration object. Using constrained optimization method, each camera coordinate system can be translated into reference camera coordinate system, and thus can get the relative positions between them. The experimental results show that the method is practical in multi view calibration.3615|253|0更新时间：2024-05-08
- 摘要：H264 is a new video compression standard with high compression rate, high video quality and high complexity. Because of the high compression rate, some normal watermarking scheme cant be used in H264 directly. In this paper, a video watermarking scheme base on the region character of motion vectors is proposed. Moreover, computational complexity is reduced by 50%. From the experimental result, after embedded the watermark, there is no difference between the original image and that of embedded watermark. The fluctuation of PSNR is controlled within ±01 and the increase of bit rate is controlled within 1%. In motion estimation, Search points are half of the original one. So this scheme is an easy and high efficient method to embed watermark. And the capacity of watermark can be doubled by judging four regions.3355|252|0更新时间：2024-05-08
- 摘要：With the rapid development of the Internet and the mobile value added services, the increasingly expansion of users and markets and the coming of Chinas 3G Time, the mobile value added services will be a bright spot to uplift the mobile communication industry. In this paper, the writer studied the security issue of transferring messages with mobile devices, and planned and realized the key management of OTA system by adding MAC (message authentication code) to ensure data security. It ensures the safety and completeness of the data transformation between the server and the USIM card; therefore it provides the guarantee to the launch of the mobile value added services.3080|214|0更新时间：2024-05-08
- 摘要：To improve the practicability of digital watermarking, watermarking imperceptibility and robustness must be ensured. Under the instruction of watermarking robust theory proposed by Cox et al, considering human vision system characteristics fully, a blind digital watermark method based on genetic algorithm is proposed. By adjusting coefficient difference based on edge detection in wavelet transformation domain, integrative optimization of watermarking imperceptibility and robustness is implemented. Experiment show that the method we proposed not only is robust to attacks, but ensures the quality of watermarking embed image.2711|239|0更新时间：2024-05-08
- 摘要：The Copy Move is a common method in digital image manipulation. In this paper, based on moment invariants, one blind detective method is investigated. The proposed method can successfully detect forged part even when the copied area is rotated, reversed and zoomed. The iterative block split is also proposed to deal with the problem of large computational complexity of the proposed algorithm. Simulation results show the efficiency of proposed algorithm.3806|230|0更新时间：2024-05-08
- 摘要：The JND (just noticed difference) threshold matrix based on block was calculated by a new human visual model, and then a text encryption algorithm based on image adaptive public watermarking technique operating in DWT domain was presented. Firstly, the text in a file was encoded with Reed Solomon codes for robust and transferred into a binary sequence, which could be permuted randomly to a final watermark. Then, two neighboring blocks were selected from the Hilbert scanning sequence of the host image blocks in turn, and transformed by 1 level DWT. By JND, the final watermark was embedded into the host image by modifying the average values of the two corresponding subbands. The experimental results show that the embedded watermark is invisible to human eyes at a low PSNR, and the algorithm is robust to common image processing operations, especially to signal enhancement. The conclusion is that the algorithm is effective and practical.3304|223|0更新时间：2024-05-08
- 摘要：To avoid poor appearance of lossless visible watermarks due to gray level flipping existence in some previous work, we propose two embedding methods capable of completely reconstructing the host image with satisfactory visual quality. The first calculates exclusive OR between the image’s MSB plane and the binary watermark. The other method uses a bit shift strategy to push down high bits of the mark covered pixels, and replaces the MSB with the mark data, making the watermark semi transparent and other image areas completely transparent. The watermark can be embedded into any position of the carrier image with any size. Both methods can losslessly reconstruct the host image without knowing the watermark information.3019|258|0更新时间：2024-05-08
- 摘要：In order to manage the road traffic in digitization and intellectualization,Vehicle Digital Standard Infosource System is proposed to establish electronic car license plate technique with the foundation of passive UHF radio frequency cards. It is explored and demonstrated in the public security road traffic management.The demonstration results indicate the system of vehicle digital standard infosource has very broad application domain and high promotion value. In the mean time,it provides the convenience to the public security road traffic digitization and the intellectualized management.3312|231|0更新时间：2024-05-08
- 摘要：With the rapid development of smart surveillance technology, more surveillance systems in public environments have switched from human based to computer. Delivery warehouse in airports is an important security area and it is very necessary to detect the inside invaders. Human detection is an important part in visual analysis and also widely used in smart surveillance systems. In accordance with the surveillance videos of airport delivery warehouse, a kind of visual feature is proposed to differentiate human body and normal luggage. Based on that feature, a real time smart surveillance system is also built to detect the invaders. This system reduces the cost of human resources and the experiment shows the missing rate of invaders detection is also reduced and the safety is increased.3519|228|0更新时间：2024-05-08
- 摘要：Due to time variable and heterogeneous network, streaming media system always can not adapt to the network bandwidth well. Based on analyzing the strategy of congestion control of Darwin Streaming Server, using TCP Westwood to estimate the bandwidth, an improved congestion control strategy is proposed in this paper and validated in a real time video transport system. Experimental result shows that the improved strategy not only improves the video quality in the receiver, but also improves the bandwidth utility.3180|246|0更新时间：2024-05-08
- 摘要：In order to protect the copyright of the vector graphics, in this paper we introduce a fusion rule for watermark detection of vector curves using SVM classification. At the embedding stage, the watermarks are embedded into multi vector curves with the same key. At the detecting stage, firstly, the watermarks are detected using the embedded key then a eigenvector is obtained which is created by detection correlative values in a certain order. Finally the SVM classification can determine whether those curves have the watermarks that embedded with a right key. The study samples of SVM classification are imitating all kinds of eigenvector, which are detected from the attacked and transformed curves. Essentially it is a fusion rule of multi related values, which based on the SVM classification. Theoretical analysis and simulation results prove it is feasible and effective.3344|225|0更新时间：2024-05-08
- 摘要：Micro doppler effect induced by non uniform motion of mobile receiver via faraway DTV transmit station was analyzed based on multipath channel propagation characteristics. To accurately separate and extract micro Doppler values of multiple signal components in DTV received signals, the support vector clustering (SVC) algorithm was applied. By using FRFT transform results of the received signal, clustering analysis was conducted, and synchronous detection and accurate estimation of multiple micro doppler components was realized. The simulation results showed that the method proposed can separate, extract and estimate micro doppler values of multiple signal components from different paths and different DTV transmit stations under single frequency network.3052|211|0更新时间：2024-05-08
- 摘要：The mechanism of patch downloading based on USIM card is proposed in this paper. This mechanism adopts authentication technology to enable the co authentication between OTA (over the air) and dynamic linking library (DLL), ensuring the mutual trust. In the paper, DLL is also suggested to practise the patch download mechanism. In this approach the application on USIM card can be updated by OTA through the control and management of patch downloading provided by the definition of lower level API.3336|242|0更新时间：2024-05-08
- 摘要：Time shifting is expected to be a popular core feature of IPTV systems because of its ability to enable users to personalize their TV channels. Here we propose a quality affordable IPTV time shifting solution to what is often the most demanding situations the combination of the limited power of end user equipments and highly interactive users on a dedicated close network. Extensive empirical evaluations based on first hand accounts of deployment experiences as well as theoretical analysis are given. Evaluation results show that the system offers good quality of time shifting service in high concurrency environment.3106|212|0更新时间：2024-05-08
- 摘要：The set top box (STB) based scalable information publication system is designed and implemented to improve publication system efficiency. This system consists of application server, central instances for STB groups and terminals. Separation of control logic and data transmission is used together with hardware redundancy and SSL to improve system scalability, reliability and usability. Data transmission is optimized to supporta large number of terminals. Evaluation result shows that this system could provide high quality information display service at a relative lower hardware cost. This system has been applied in retail stores, restaurants in Shanghai, Beijing and Shandong.3122|243|0更新时间：2024-05-08
- 摘要：The mobile streaming media application is an application of high data bandwidth which is the most valuable new service among the 3G network based services.In order to further promote the mobile streaming media technical application, 3G mobile network and fixed network video surveillance is combined to realize both advantages. Further more, a mobile video surveillance system prototype based on a 3G network is proposed. This model can make up for the traditional fixed line and mobile networks in the problem of insufficient bandwidth, the mobile video surveillance system design have a certain significance.3638|232|0更新时间：2024-05-08
- 摘要：In order to facilitate video production of the front end, a smart displaying scheme for real time soccer video on portable devices was proposed. Using the proposed scheme, which can solve the problem of wonderful specifics loss under the reduced resolution users can appreciate real time soccer games comfortably under low resolution circumstance of portable devices. Different from typical football processing techniques, the proposed scheme exploits a wholly region based approach to analyze real time soccer video, and can reduce the computational complexity substantially and satisfy the real time processing requirements while loosen the dependency on hardware equipments. The proposed scheme is composed of two stages: shot classification and dynamic extraction and playing of region of interest (ROI). The experimental results manifest that the scheme is indeed efficient, practical as well as reliable, and can play real time soccer video smartly on portable devices.2929|237|0更新时间：2024-05-08
- 摘要：The H264 video compression standard is extensively applied thanks to its excellent compression efficiency and coding flexibility. A moving object segmentation approach in H264 compressed domain is proposed in this paper. The motion fields are first extracted from the compressed video, in which the noise vectors are removed by weighted median filter. Then the predicted motion fields reconstructed by backward estimation are used to accumulate the motion field. After that, the modified statistical region merging is exploited to segment the moving object based on three motion characteristics magnitude, divergence and curl. Experimental results demonstrate that our approach can efficiently extract the moving objects from H264 compressed and as the segmentation quality is good.2783|229|0更新时间：2024-05-08
- 摘要：Inter frame mode decision is the most time consuming part in H264. A novel fast mode decision algorithm based on all zero block and consistency of moving direction is proposed in this paper. Firstly, the sufficient condition of absolute all zero block and the necessary condition of relative all zero block are defined to divide the MB into four classes. Secondly, in each class, the moving state of neighboring MB and the SKIP RD cost of this MB are used to further classify the MB into some categories so that the range of candidate modes can be narrowed. The experimental results have shown that the proposed algorithm can speed up inter frame mode decision greatly and incurred loss in PSNR and increment in rate can almost be neglected.2809|203|0更新时间：2024-05-08
- 摘要：In this paper, a motion retrieval and recognition system is investigated from a ensemble learning model. In order to recognize and retrieve 3D motion data, first motion features are extracted from motion data. Due to the high dimensionality of motion’s features, a generalized isomap nonlinear dimension reduction based on the estimation of underlying eigenfunction is used for training data of ensemble HMM learning. Then each action class is learned with one HMM. Since ensemble learning can effectively enhance supervised learning, ensembles of weak HMM learners are built. Experimental results show that our approaches are effective for information retrieval from large scale motion database.2872|217|0更新时间：2024-05-08
- 摘要：Frame rate up conversion is used in conversion between any two display formats, low bit rate video communication and many other fields. A novel adaptive frame rate up conversion algorithm is presented in this paper. The variable size block motion compensation based on pre knowledge is incorporated into bilateral motion estimation and the adapted multi stage motion vector processing and adaptive motion compensated interpolation are adopted so as to resolve the problems of overlapping, hole and block artifacts as well as guarantee higher computation speed. Experimental results show that the proposed algorithm provides better image quality than conventional methods both objectively and subjectively, and the computation efficiency is also improved.3591|220|0更新时间：2024-05-08
- 摘要：Because a linear model is usually used to predict the MAD and the QP quantization is the same in the JVT G012, which costs too much in operation and cant avoid deviation, an improved rate control algorithm for the basic unit layer enconder is presented. Firstly, the MAD is predicted by a new weighted time and space model. Secondly, macroblock layer rate control strategy is used to distribute target bits. Finally, the quadratic RQ model in the JVT G012 is still adopted to calculate QP. Simulation results show that the PSNR is effectively elevated and lower bit rate is achieved in the H.264 encoder by using the new algorithm.3308|235|0更新时间：2024-05-08
- 摘要：The request for efficient compression of hyperspectral images becomes pressing. A new lossless compression algorithm based on 3D contexts prediction for hyperspectral images is presented. Spectral band grouping algorithm is introduced to divide hyperspectral images into groups according to the neighboring band correlations, then band reordering is performed for each group. The important bands containing large information can be determined by using adaptive band selection algorithm, on which clustering is carried out according to the spectral vectors. 3D contexts are defined based on the neighboring causal pixels in current band and the corresponding colocated causal pixels in reference band. Combined with the clustering results, the optimal predictive coefficients of each cluster are trained respectively. Experimental results show that the proposed algorithm can give better lossless coding performance.3265|218|0更新时间：2024-05-08
- 摘要：A high throughput inverse transform IP core for H.264/AVC was proposed in this paper. The improved T architecture was presented to synchronize three different transforms and inverse ZigZag scan module. By applying time multiplexing buffer management to inverse Hadamard transform, we efficiently reduce its latency. Separability property of IDCT is also utilized to minimize its area. At last, the results of synthesis are given with Xilinx Virtex2 while XC2V6000 as the target device. The simulation performance shows that the design can effectively support the real time decoding of 1080i 50Hz HD stream.2923|215|0更新时间：2024-05-08
- 摘要：H.264 introduces intraprediction coding technique to improve the intracoding efficiency.This enormously increases the computational complexity. In order to decrease the computational complexity of the algorithm, a fast algorithm for intra prediction mode decision based on pixel’s edge vector’s variance, the relation between subblock’s mode and macro block’s mode. Experimental results show that compared with the original intra prediction coding scheme, the proposed algorithm reduce the computational complexity of intra prediction coding dramatically with only trivial loss of PSNR and increment of bite rate.3358|220|0更新时间：2024-05-08
- 摘要：A new method for geometric construction of QC LDPC is proposed. The parity check matrix constructed through this method has a girth more than eight, which promises good performance. Since the parity check matrix is composed of cyclic submatrices, the hardware encoder is simple. Simulations show that this type of LDPC codes has remarkable performance with low encoding complexity.3528|221|0更新时间：2024-05-08
- 摘要：A hidden Markov model (HMM) has good adaptability, can automatically learn and adapts well when used in predicting temporal data. Changing of scenes in soccer video is a fundamental trait and manifests semantics of soccer video. In this paper, a new method based on scene classification and HMM is given to analyze semantics of video, detecting semantic events in soccer video. In order to solve the low accuracy and low robustness of scene classification based on traditional method, model of visual attention and gist of scene is used in classification of scenes of soccer video. The experimental results show higher efficiency and higher accuracy of our method in detecting free kick in soccer video.3333|229|0更新时间：2024-05-08
- 摘要：In this paper, a content based image retrieval algorithm using color edge histogram in HSV color space is proposed. Firstly, the color edge is extracted by using Canny detection operator. Secondly, the color edge is transformed from RGB space to HSV space and quantized. Thirdly, the color edge is divided into some ring regions and angle regions, and the color histogram of each ring region and angle region are computed as image features. Finally, the similarity between color images is computed by using a combined feature index based on the color histogram of ring region and angle region. Experimental results show that the proposed image retrieval is more accurate and efficient in retrieving the user interested images.3879|263|0更新时间：2024-05-08
- 摘要：To measure perceptual similarity between an original image and its modified version, we propose an objective measure that is relatively stable to normal image processing but quite sensitive to significant changes of the image content in local areas. This is achieved by low pass filtering the two images, dividing them into overlapping blocks and determining similarity between the corresponding blocks in terms of correlation coefficient that is mapped to the interval \[0, 1\]. Based on previous calculated correlation coefficients, a ratio is calculated between the smallest and largest correlation coefficients and defined as the perceptual similarity. 〖BP(〗Products of a predefined number of the smallest and largest correlation coefficients are calculated. Perceptual similarity is defined as the ratio between these two products.〖BP)〗 Experimental results show that the proposed metric is not substantially affected by normal image processing. It provides indication of changes in the image contents when its value becomes less than a given threshold. The proposed metric is useful in applications such as image hashing and CBIR.3223|199|0更新时间：2024-05-08
- 摘要：For reducing the errors which arisen during Hamming error correcting decoding so that the likelihood of mistakenly and randomly rectifying is cut down and the quality of restored image is improved,an improved algorithm of restoring image based on Hamming error correcting codes is presented. This algorithm first divides the most significant bit (MSB)plane of a grayscale image into four big blocks.Then each big block is divided into four small blocks and they are exchanged vertically in a big block.Secondly,the four binary bits at the same pixel position of four big blocks are encoded according to (7,4) Hamming encoding and the three checkout bit plane big blocks are implemented with diagonal exchanging in itself,then the exchanged three big blocks are embedded into the least significant bit(LSB) of corresponded pixel of the first three big blocks of the original image orderly according to checkout bit sequence.Then the second most significant bit plane(2MSB) and the third one(3MSB) are operated using above algorithm and embedded respectively into the second least significant bit plane(2LSB) and the third one(3LSB). The experiment results indicate that the improved method has obtained better restored image effect than the original one in the circumstance of cutting,replacing,average filtering and median filtering, and JPEG lossy compression.And the noise caused by the restored algorithm is reduced largely.关键词：most significant bit(MSB);Hamming error correcting encoding;exchange;image restoration3707|224|0更新时间：2024-05-08
- 摘要：Stereo video object segmentation is a key technology in the multiview video interacting. In this paper, based on the compressed domain information a novel stereo video object segmentation algorithm is proposed to improve the validity and accuracy of the segmentation. Firstly, we extract motion vectors and disparity field to segment the initial video object. Then we used the mean shift algorithm to cluster the video object. Experimental results shows that the proposed algorithm could obtain the object with semantic confornity from the scene with complex texture.关键词：stereo video;motion vectors;disparity;mean shift6899|226|0更新时间：2024-05-08
- 摘要：The classic fast search algorithm can effectively reduce the volume of search, but because of their selective searching it will easily fall into the local optimal error. The Multiresolution Motion Estimation (MRME) and the Wavelet Multiresolution Motion Estimation (WMRME) can reduce the local optimal error to a certain extent, but they also bring a corresponding increase in computing capacity. Based on the Wavelet Multiresolution Motion Estimation, a Wavelet Leapfrog multiresolution Motion Estimation (WLMRME) algorithm is proposed.Combined with a rapid search algorithm,WLMRME can achieve better coding effect with less complexity than WMRME.关键词：wavelet domain;multi resolution;movement estimates3520|215|0更新时间：2024-05-08
- 摘要：This paper introduces the design and implementation of a semantic concept based video retrieval system, which consists of shot boundary detection and key frame extraction subsystem, semantic concept detection subsystem and user retrieval subsystem. First, digital video is divided into hierarchical structure for retrieval. Then, efficient low level feature of key frames are extracted. Support Vector Machine is used to detect concepts in such key frames, and the video retrieval is based on those concepts. In the procedure of concept detection, we take a linearly weighted fusion method based on validation precision to improve the average precision. Experiments show that the Mean Average Precision of our system is as high as the best one of all submissions.关键词：video retrieval;support vector machine (SVM);concept detection;fusion3359|261|0更新时间：2024-05-08
- 摘要：Active Appearance Models (AAMs) is an effective statistical method to build 2D model for an object, which combines shape and texture information.A novel method for building 3D face model is proposed, which makes use of the depth information and corresponding intensity information generated by correlation image sensor (CIS), and extends 2D AAMs to 3D AAMs. The proposed improved AAMs fuse the shape, texture and depth information of face to build 3D face model. In facial recognition experiments which using 3D facial images based on CIS imaging system, the improved 3D AAMs model shows better recognition result than traditional AAMs algorithms and Eigenface.关键词：facial recognition;CIS (correlation image sensor);3D facial imaging;3D AAMs(active appearance models)3715|242|0更新时间：2024-05-08
- 摘要：In this paper, a new trellis coded watermarking algorithm of trellis codes combine with constant robustness embedding is proposed. In the algorithm, both informed coding and informed embedding are used. The watermark is encoded by a dirty paper trellis encoder. Then the strength of the noise to be added to reduce the normalized correlation coefficient to a given threshold is computed, which is taken as a constant robustness parameter. The watermark signal is embedded into the cover work by Gram Schmidt orthogonal algorithm under the condition of the constant robustness parameter, and Viterbi algorithm is used to detect and extract the watermarked signal. Experimental results indicate that the proposed algorithm can improve the robustness to white Gaussian noise and amplitude scaling.3207|231|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