基于划分-合并和稳健M-估计的运动图像分层表示方法

LAYERED REPRESENTATION OF IMAGE MOTION USING SPLIT-AND-MERGE METHOD AND ROBUST M-ESTIMATION

本文提出了一种基于划分-合并和稳健M-估计以及最小描述长度约束的运动图像分层表示方法．与现有的方法比较，算法在参数估计方面有较大改进，不仅免除了人工干预，而且计算量也比较小．在参数优化过程中，采用了更符合实际情况的主成分滤波，得到了满意的结果．

A method for layered representation of image motion using split-and-merge method and robust M-estimation combined with MDL restriction is presented. The algorithm introduces better parameter estimation methods that works without human interaction and is easier to optipute compared with known methods. Another improvement is adoption of a majority filter that gives more accurate results according to actual requirements in stage of layered parameters' optimization.