The Research Of Fast Motion Estimation Based On H.264/AVC

H.264/AVC is a new generation of international video codingstandard jointly developed by ITU-T and ISO/IEC. Compared with theprevious video standards, with the advanced nature of the technicalaspects, open of research and application of international and othercharacteristics, giving rise to widespread concern and research indomestic and foreign experts and scholars. However, the existingresearch results show that: the high performance of H.264/AVC is thecost of high complexity, which lead to the computational efficiency ofthe entire video codec significantly. This can not maximize thesatisfaction of the demand for real-time video, a serious impediment tothe promotion and application in real-time video. The motionestimation technology is the key technology in the video codec, butalso the most time-consuming module. Therefore, improving the searchefficiency of motion estimation has become the key to improve theperformance of video codecs. At present, under the premise to ensurethe accuracy of video coding, maximize the search speed of motionestimation as the current video field of research focus.All along, distinguished scholars and experts in order to improve theefficiency of motion estimation search made unremitting efforts, hasmade a three-step method (TSS), hexagon search (HEXS), diamondsearch algorithm (DS), UMHexagons algorithm, EPZS algorithm fastmotion estimation algorithm. The UMHexagonS algorithm has absorbedthe advantages of a variety of motion estimation algorithm framework,which can maintain a good rate-distortion performance and videoquality, improve video coding efficiency, which has been widespreadconcern. Therefore, this paper take the UMHexagonS algorithm as thestarting point, the principle and motion estimation motion estimationalgorithm on the basis of the key technologies of various improvements in the study of motion estimation, trying to achieve a breakthrough inimproving the motion estimation algorithm performance.In this paper, the first describes the H.264/AVC overall architectureand key technologies, and then analyze the basic principles of motionestimation and several classic fast motion estimation algorithm. Finally,in-depth study on UMHexagonS algorithm based on the proposedimproved algorithm. In particular, take full advantage of thecharacteristics of the video sequence itself (movement characteristics,texture features) and the law of motion of the object on the basis offour parts UMHexagonS algorithm has been improved.1. Matching criteria, Summing the absolute error criterion (SAD)matching templates using a uniform square template and the objectitself does not match the law of motion, and contrary to the things ofthe unified law of diversity. Law of motion improved to match thecriteria for fully integrated with the natural objects, making thematching template to select the more scientific.2. Predicted search starting point, First according to the temporalcorrelation of motion vectors to determine the type of movement of thecurrent block; then choose a different prediction algorithms for differenttypes of movement, if the current block is a small movement of thetype,(0,0) for the search starting point, if the currentblock themovements larger type, according to the spatial correlation of motionvectors to predict the search starting point.3. Search_range, Integrated to determine the dynamic searchrange, the range of motion of the video image, the search scope of theprinciple of selection and the current prediction block type three togreatly improve the search speed. In order to ensure the accuracy ofvideo encoding, consider the minimum search window to avoid fallinginto local minimum.4. Search strategy, According to the spatial correlation between the pixels, the early termination of the algorithm is designed based onthe statistical results; cross-bias distribution of features and texturefeatures in the current block according to the motion vector of thecenter-the adaptive improvement of the search template.Finally, improvements in several aspects of this technology into theoriginal UMHexagonS algorithm, select a variety of test sequences tothe overall improvement of the algorithm performance tests, resultsfrom a comprehensive and objective data and subjective analysis ofthe improved performance of the algorithm. Improved algorithm withthe original UMHexagonS algorithm, the video quality is almost notaffected under the premise of greatly reducing the coding complexity,greatly improving the coding efficiency. Thus, it can effectively meetthe needs of real-time video, and can be used in pattern recognition,video retrieval and intelligent surveillance video.