为了降低多视点视频编码（MVC）的计算复杂度,提出了一种基于运动矢量视点间相关性和空间相关性的多视点视频编码快速运动估计算法.该算法从邻近视点的运动矢量中获得反映真实运动状态的参考运动矢量,从参考运动矢量以及空间相邻运动矢量中选取搜索中心,并利用搜索中心与参考运动矢量的偏离程度自适应预测搜索范围,以达到降低运动估计复杂度的目的.实验结果表明,与MVC参考代码中的全搜索算法和快速搜索算法相比,采用文中算法分别减少了95.6%和22.2%的编码时间,同时保持了编码率失真性能.

In order to reduce the computational complexity of multi-view video coding（MVC）,a fast motion estimation algorithm is proposed based on the inter-view and the spatial correlations of motion vectors.In this algorithm,first,the reference motion vector,which accords with the real motion state,is obtained from the motion vectors in neighboring views.Then,the search center is selected from the reference motion vector and the spatial neighboring motion vectors.Finally,the deviation of the search center from the reference motion vector is utilized to adaptively predict the search range.Thus,the computational complexity of motion estimation is effectively deduced.Experimental results show that,as compared with the full search algorithm and the fast search algorithm in the MVC reference software,the proposed algorithm reduces the encoding time by 95.6% and 22.2%,respectively,with the rate distortion performance keeps unchanged.