摘要
文章提出了一种基于离散粒子群优化算法的块匹配运动估计算法。该算法将块匹配运动估计的局域性搜索与离散粒子群算法的全局性搜索结合起来,并针对运动矢量的特点,采用了Gray码编码、运动矢量预测以及有效的迭代提前终止准则等策略,克服了以往快速搜索算法容易落入局部最优的问题,在获得与全搜索算法相近的搜索精度的同时,降低了平均搜索点数。实验结果表明,对于运动复杂度较高的序列,该算法仍能保持较好的性能。
In this paper, a block matching algorithm（BMA） based on discrete particle swarm optimization（DP-SO） is proposed for motion estimation. By integrating the local searching of BMA with the global searching of DPSO and using strategies like Gray encoding, motion vector prediction and effective early termination criteria of iteration according to the features of motion vector, the proposed algorithm overcomes the shortcoming of being liable to local optimum from which traditional algorithms often suffered. The simulation results show that the proposed algorithm obtains almost the same accuracy as full search algorithm with fewer search points and performs well in treating video sequences with high motion complexity.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2011年第11期1661-1665,共5页
Journal of Hefei University of Technology：Natural Science
基金
高等学校博士学科点专项科研基金资助项目(20060359004)
国家科技部中小企业创新基金资助项目(09CZ6213401392)
关键词
运动估计
块匹配算法
离散粒子群优化算法
运动矢量
motion estimation
block matching algorithm（BMA）
discrete particle swarm optimization （DPSO）
motion vector
