基于伪树冠的细粒度可变尺寸块运动估计

为了实现细粒度的可变尺寸块运动估计算法,构造了一个三叉和四叉结构组合的伪树冠数据结构,和细粒度四叉树森林结合成一个具有单个根节点的复合树。细粒度可变尺寸块运动估计算法同粗粒度算法一样可以利用树结构的特点快速找到具有率失真斜率极值的节点。仿真试验验证了细粒度算法和粗粒度算法的不同适用范围。通过仿真试验分析了运算量减少策略、搜索窗大小、搜索深度对算法性能的影响。

Fake Tree-crown Based Fine Granularity Variable Size Block Motion Estimation

A fake tree-crown data structure is introduced to implement fine granularity variable size block motion esti mation algorithm(FG-VSBM).By combining fake tree-crown and fine granularity quad-trees into one tree with single root,max or min rate-distortion slope node finding scheme is transplanted from coarse granularity variable size block motion estimation algorithm(CG-VSBM)to FG-VSBM.The simulation experiments exemplify the different application scope of FG-VSBM and CG-VSBM.The influence of computation reduction scheme ,size of searching window and depth of constrained searching to VSBM algorithms is analyzed in detail from the experiments results.