运动估计中UMHexagonS的研究与改进

Research and improvement on UMHexagonS for motion estimation

针对UMHexagonS算法搜索点数较多，运算量较大以及耗时等问题，提出了改进方案。一方面，对部分搜索模板提出了一种菱．八边形搜索模板和一种多层次八边形搜索模板；另一方面，在整个搜索过程中，结合象限区域分割法，可以有效减少搜索点数和搜索时间。实验结果表明，改进后的算法在保证较好的PSNR和码率情况下，比原算法减少了4％～26％的运动估计时间。

As the UMHexagonS algorithm exists problems such as excessive searching points, complicated computation and consuming time etc., this paper puts forward some improving plans. On the one hand, to some of the searching template, a diamond-octagon searching template and a multi-level-octagon searching template are proposed; on the other hand, in the whole searching process, when using quadrant regional segmentation method, searching points and searching time can be effectively reduced. The experimental results show that the improved algorithm in the situation of guaranteed Peak Signal to Noise Ratio （PSNR） and bit rate, compared to the original algorithm, saves the 4%-26% of the motion estimation of time.