作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2007, Vol. 33 ›› Issue (09): 210-211,.

• 人工智能及识别技术 • 上一篇    下一篇

基于图像特征的Winner-update运动估计策略

李 军,李艳辉   

  1. (暨南大学珠海学院计算机系,珠海 519070)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-05-05 发布日期:2007-05-05

Image Feature Based Winner-update Strategy for Motion Estimation

LI Jun, LI Yanhui   

  1. (Department of Computer, Zhuhai School, Jinan University, Zhuhai 519070)
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-05-05 Published:2007-05-05

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。

关键词: 运动估计, 快速算法, 块匹配

Abstract: 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%.

Key words: Motion estimation, Fast algorithm, Block match

中图分类号: