基于图像特征的Winner-update运动估计策略  

Image Feature Based Winner-update Strategy for Motion Estimation

作　　者：李军[1] 李艳辉[1] 

机构地区：[1]暨南大学珠海学院计算机系,珠海519070

出　　处：《计算机工程》2007年第9期210-211,221,共3页Computer Engineering

基　　金：暨南大学引进人才基金资助项目(04JZKY004)

摘　　要：运动估计是视频压缩算法的重要组成部分,块匹配算法是运动估计的主要方法,但算法复杂度高。该文提出了一种Winner-update搜索策略,总结了衡量图像复杂度的简便方法,构造合理的匹配误差序列,以重排匹配搜索的顺序,进一步降低了块匹配算法的计算复杂度。文中设计了与算法相适应的数据结构,对比实验表明了该算法可节省90%～98%的计算量。The block match algorithm, as an important part of video compression, is the most popular for motion estimation, but its computational complexity is high. A fast algorithm for motion estimation using winner-update search strategy is presented in the paper. A simple method is proposesd to measure the complexity of the image, and the proper matching distortion set is constructed. The new algorithm reduces further the computational complexity of block matching by rearranging the searching order according to the image complexity. A corresponding data construction is designed for the new algorithm, and experimental results verify the effect of the new algorithm, it saves computations about 90%～98%.

关 键 词：运动估计 快速算法 块匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]