在极低编码速率条件下，Neff和Zahor提出的基于匹配跟踪信号分解的视频编码器不仅具有比H．263编码器更高的编码性能，而且能够避免产生人眼敏感的方块效应，但由于该算法需要在一个冗余字典里搜索最佳匹配误差结构的原子函数，其实现所需要的运算量比传统的编码器要高很多，因而影响了该编码器的效率。为了提高编码效率，在对能量优先原子搜索策略进行分析的基础上，提出了一种改进的全搜索策略和加权能量优先搜索策略，从而改进了最佳时频原子搜索策略。最后还对搜索策略的编码性能和运算效率进行了评价和实验。

At very low bit rates,the video coder based on matching pursuits proposed by Neff and Zahor can avoid producing visual-sensible block artifacts and has been proved to have better coding performance, compared with H.263 standards. But, the computation complexity of its realization is much too higher than traditional coders, because it needs to search optimal matching atoms in a redundant dictionary,and the searching burden strongly affects the efficiency of the coder. In order to advance the efficiency of traditional coders, The author first analyzes the atoms searching scheme based on energy priority, then develops a improved full searching scheme and a weighted searching scheme based on energy priority, Both novel time-frequency atoms searching algorithms improve the optimal time-frequency atoms searching scheme. At the same time, we compare and experiment the coding performance and computation efficiency of our algorithms.