DVC边信息估计中块匹配准则算法的研究

摘要

The block matching algorithm which is usually adapted in using side information estimation by using interpolation method, is analyzed in detail, and its existing problems are also pointed out. The problems are that when high motion occurs or longer step lengths are selected,the popular MAD criteria is used to minimize the individual block distortion results in a quite noisy motion vector field. So in order to improve the precision of the motion vector, a new matching estimation algorithm is proposed in this paper which considering the cost of the motion vector by a smooth coefficient, and the experiment results show that this method can remove the image square effect effectively and can improve the quality of the side information.%对在采用插值法生成边信息的过程中通常用到的块匹配算法进行了分析,指出了常用算法存在的问题即对于运动比较剧烈的视频序列或采用的搜索步长比较大的情况下,通常的匹配准则方法估计出来的运动矢量与实际的运动矢量之间存在较大的偏差,因此为了提高运动矢量的精度,采用一个平滑系数将运动矢量的代价值应用到加权SAD中作为新的块匹配的估计准则,实验结果表明,该方法能有效去除图像的方块效应并能提高边信息的质量.