欢迎访问《图学学报》 分享到:

图学学报

• 视觉与图像 • 上一篇    下一篇

引入搜索预测与阈值决策的改进菱形运动估计算法

  

  • 出版日期:2015-08-28 发布日期:2015-08-04

Improved Diamond Motion Estimation Algorithm Based on Search Prediction and Threshold Decision

  • Online:2015-08-28 Published:2015-08-04

摘要: 为了提高视频的压缩效率,在传统菱形搜索算法基础上提出一种改进菱形搜索算
法。该算法通过引入动态阈值,在起始搜索点预测、菱形搜索模式和搜索中止算法方面进行了
优化,减少了SAD 计算的内部冗余和搜索区域中不相关的块匹配计算,同时采用自适应搜索模
式选择技术减少运输复杂度。实验结果表明:提出的改进菱形搜索算法适合各种运动类型的视
频序列,特别适用于运动变化剧烈的序列,相比于FS 算法,能够在PSNR 值和码率值极其接
近于FS 算法的情况下对所有序列的MET 减少约95%,大大减少运动估计时间。

关键词: 视频压缩, H.264, 运动估计, 菱形搜索, 块匹配

Abstract: To improve the compression efficiency of the video, a self-adaptive diamond search
algorithm is put forward based on traditional diamond search algorithm. The algorithm is improved in
the prediction of the beginning search spot, diamond search mode and search suspended algorithm by
bringing in dynamic threshold. It realizes the self-adaptive search, which reduces the internal
redundant SAD operation and skips all the irrelevant blocks in the search area. The experiment result
shows that the self-adaptive diamond search algorithm suits all kinds of motional video sequence,
especially those sequences changing poignantly in movement. Comparing to the FS algorithm, the
improved algorithm decreases approximately 95% of the motion estimation time (MET) of all the
sequences under the condition that the PSNR and the code rate value are very close to FS algorithm.
The motion estimation time is greatly decreased.

Key words: video compression, H.264, motion estimation, diamond search, block matching