基于数学形态学的运动区域估计及运动图象编码

MATHEMATICAL MORPHOLOGY-BASED MOTION ESTIMATION AND VIDEO CODING

提出了基于数学形态学的运动区域估计和运动图象编码新算法．这种算法利用数学形态变换提取帧间物体运动块和运动区域，并进行自然形状对应的区域匹配和运动补偿．实验表明，与ＭＰＥＧ等分块运动补偿相比，本算法运动匹配准确，搜索快速，避免了方块效应，在相同码率下较好地改善了图象质量，特别适用于低码率运动图象编码．

This paper presents a new algorithm for motion estimation and video coding which is based on the mathematical morphology. The proposed algorithm, by using morphological set operations, segments image objects and extract the motion regions, and makes the region matching and motion compensation according to the image objects'particularities. The experimental results show significant improvement in eliminating the block artifacts. Compared with MPEG motion compensation, the algorithm can improve the predicted image quality at the same bit rate and is especially suited to the low bit rate video coding.