基于运动对象检测的UMHexagonS搜索算法改进

An Improvement of UMHexagonS Based on Detecting Moving Object

在H．264精细可分级编码中，针对减少运动估计时间，本文提出一种基于运动对象检测的UMHexagonS搜索算法改进．首先，根据运动对象检测区分出运动区域和背景区域，对不同的区域采用不用的搜索算法；其次，利用两次菱形模版替换原来的5×5螺旋搜索，减少了50％的搜索点数，且能提前结束搜索．实验结果表明，改进后的算法，在保证图像质量基本不变的前提下，运动估计时间减少10％～40％．

Abstract： In FGS video coding, in order to reduce the time of motion estimation, this paper proposed an improvement algorithm of UMHexagonS, which is based on detecting moving object. First, distinguish between motion region and background according to the motion object detection, use different search algorithm in different areas. Then, uses two diamond template replace the original 5 × 5 spiral search, decreases 50% of search points, can terminate search early. The experiment results show that the improved algorithm in the premise of image quality can reduce 10%-40% of the motion estimation time.