论文部分内容阅读
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．
A motion estimation method based on regional motion correlation is studied. In the block matching motion estimation, the block with good motion correlation is predicted based on the motion of the neighboring block or the previous frame, and then it is searched in a small area to optimize the motion vector. The traditional global search block matching motion estimation algorithm is given to improve the consistency of motion vector correction criteria. According to the motion estimation error of the previous frame, the algorithm decides whether to predict the motion of the current block and avoids the spread of motion estimation error. The experimental results show that the motion vector consistency obtained by the proposed algorithm is much higher than that obtained by the traditional global search block matching method, while maintaining the prediction quality comparable to the global search method, while the computation time is greatly reduced.