论文部分内容阅读
研究了一种基于时空相关性的运动估计方法。提出在块匹配法中,对于运动相关性好的块根据相邻块或前一帧图像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。对传统的块匹配运动估计,给出了提高运动矢量一致性的修正准则。实验结果表明,本文算法所得的运动矢量一致性远高于传统块匹配法,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缩短。
A motion estimation method based on spatiotemporal correlation is studied. In the block matching method, for the block with good motion correlation, the initial motion vector is predicted based on the motion of the neighboring block or the previous frame, and then search in a small area to optimize the motion vector. For the traditional block matching motion estimation, the correction criterion for improving the consistency of motion vectors is given. The experimental results show that the motion vector obtained by this algorithm is much better than the traditional block matching method, while maintaining the quality of prediction comparable with the global search method, while the computation time is greatly shortened.