The Research Of Fast Motion Estimation And Lens Boundary Detection

Internet is developing rapidly. In order to ensure the efficiency and real-time ofvideo datas transmission, we must encode the datas according to certain videocompression standards. As the core step in the process of video encoding, motionestimation has great influence on the efficiency of video compression and encoding.Its time complexity is the highest in encoding. As the times goes by, the traditionalvideo retrieval technology will be out of style. The video retrieval technology basedon details will emerge and the retrieval technology based on the compressed domainin content retrieval will be concerned widely.The paper is on the basis of the knowledge above. At first, we analyse thebackground and significance of argues and review the development history of videocompression standards. Then we introduce the research status of motion estimationand the realized value of Lens boundary detection. Moreover, we summarize theproblems to study and the main clues in it.By introducing some unnecessary search points in the searching process in theUMHexagonS algorithm, we put forward an improvement scheme. In the scheme, itrefines the search area and adopts two kind of forecasting method about initial points.According to the effect on predictions by the initial points, we reduce the searchpoints and optimize the search path owing to different search strategy.At last，we study the boundary detection about video lens. The motion vector gotfrom the motion estimation in the compression area can reflect the behaviorcharacteristics of the video objects. It will be focused that detecting the shot boundaryby draw key messages. So, basing on the MPEG video streaming fast-divisionalgorithm, we propose an improved scheme. In the scheme, we can detect shotboundary by draw the message of the motion vector. It can achieve betterperformance.