摘要
为了提高基于块匹配的运动估计的速度和精度 ,提出了一种带中心偏置点检测模式的自适应快速块运动估计算法 .该算法根据图像序列的运动向量基于中心偏置分布的特点和相邻块运动向量间的高度相关性 ,依据块的不同运动内容来确定其搜索起点、搜索范围和搜索策略 ,从而实现块运动向量的快速而有效地估计 .同时 ,对于大运动块 ,采用了多候选者方式 ,进一步提高了搜索精度 .实验结果表明 ,该算法的搜索速度接近 N3SS,N4SS,而搜索精度比它们高 ,与 HSS相似 ,接近 FSBM.
In order to make block motion estimation faster and more accurate, a fast, efficient and adaptive block motion estimation algorithm with center-biased pointer-detect mode is described in this paper. Based on the characteristic of center-biased distribution of image sequences' motion vectors and high correlation of adjacent blocks' motion vectors, the algorithm achieves block motion estimation fast and efficiently by adaptively determining search origins, search ranges and search measures of different blocks. At the same time, the motion estimation accuracy is improved when multi-candidate mode is applied to those blocks with large displacements. The simulation results show that the searching speed of our algorithm is close to that of N3SS and N4SS, and the matching accuracy is similar to that of HSS and close to that of FSBM.
出处
《计算机研究与发展》
EI
CSCD
北大核心
2001年第6期679-683,共5页
Journal of Computer Research and Development
基金
国家"八六三"高技术研究发展计划基金资助! (863 -3 17-0 1-10 -99)