区域运动相关性的运动估计方法

研究了一种基于区域运动相关性的运动估计方法．提出在块匹配运动估计中，对于运动相关性好的块根据相邻块或前一帧图象的运动预测出它的初始运动矢量，然后在小范围内搜索，做运动矢量的优化．对传统的全局搜索块匹配运动估计算法，给出了提高运动矢量一致性的修正准则．算法根据前帧运动估计误差来决定是否对当前块运动作预测，避免了运动估计误差的扩散．实验结果表明，文中算法所得的运动矢量一致性远高于传统全局搜索块匹配法所得的结果，同时能保持和全局搜索法相媲美的预测质量，而运算时间则有较大幅度减小．

Motion Estimation Based on Region Coherence

A new motion estimation algorithm based on spatio temproal coherence is presented, in which an initial motion vector of the block having good coherence with its neighbors or that of the previous frame is directly predicated. A more precise motion vector is determined based on the initial one by full searching BMA in a smaller range. For the blocks with poor coherence, full searching BMA with an ordinary range is used, and their motion vectors are at last adjusted by the coherence criterion given in the paper. The motion compensation prediction error of the previous frame is used to determine whether or not to predict the initial motion vector of the current block, and thus it avoids spread of the prediction error. The experiments show that satisfying results are reached with the proposed algorithm.