摘要
为了提高视频编码效率,提出了一种基于变异粒子群的快速运动估计算法。新算法将运动矢量特性和粒子群算法的全局搜索特性结合,通过迭代寻找全局最优解。为了避免粒子群算法陷入早熟收敛,在每次迭代中加入了变异算法。同时采用合适的终止策略降低运算复杂度。实验结果表明,新算法对运动平缓的视频序列,搜索精度和运算复杂度与以往快速搜索算法相当。对运动中等和剧烈的视频序列,新算法的搜索精度比以往快速搜索算法可提高0.2~1.7 dB,并能减少视频质量大的波动,运算复杂度略高于以往快速搜索算法,但不到全搜索算法的4%。
In this paper,a fast motion estimation algorithm based on particle swarm optimization（PSO） with mutation is proposed to improve the video coding efficiency.New method combines the characteristic of motion vector and PSO and searches global optimum solution through iteration.Mutation is added in case of premature convergence of PSO.Also new method adopts termination strategy in order to reduce computational complexity.The experimental results show that the search precision and computational complexity of new method is similar to the exiting fast search algorithms for the video sequences with slow motion.For the video sequences with middle and violent motion,the search precision of new method has higher 0.5~1.7dB and reduces the significant fluctuation of video quality,the computational complexity of new method is slightly greater than previous fast search and much less 10% than full search.
出处
《电子测量与仪器学报》
CSCD
2011年第1期23-28,共6页
Journal of Electronic Measurement and Instrumentation
基金
中央高校基本科研业务费专项资金(编号:ZYGX2009J024)资助项目
关键词
视频编码
运动估计
粒子群算法
变异算法
video coding
motion estimation
particle swarm optimization
mutation
