一种新的基于遗传算法的快速运动估计方法

A Novel Fast Motion Estimation Method Based on Genetic Algorithm

本文提出了一种新的基于遗传算法的快速运动估计方法 .该方法对遗传算法进行了改进 ,采用“阈值法”确定选择算子 ,并将基因变异所导致的随机搜索与特定目标搜索相结合 ,解决了以往快速搜索算法易陷于局部最优的问题 ,大大提高了运动估计速度 .该方法还将运动矢量空间一致性原则用于初始种群的选取 ,进一步提高了算法性能 .由于其具备遗传算法固有的规则性和高度并行性 ,该方法适合于采用VLSI实现实时视频编码器 .

A novel method of motion estimation based on an improved genetic algorithm is presented.By applying a threshold selection operator and combining objective search with random search derived from genetic mutation,the method can overcome the shortcoming of being liable to local optimum,thus the estimation speed is prompted.The performance of search is also improved by selecting the initial population based on the coherence rule within the motion vector space.The inherent regularity and high parallelism of genetic algorithm make it feasible for VLSI implementation of video encoders.