Abstract: Block-matching algorithm (BMA) is always the focus of video coding techniques because of its simplicity and efficiency. In the past decades, many fast BMAs are proposed to realize the goal of coding video in time. Among them, the fastest one is the OTA (one at a time algorithm), which is very simple, fast and effective with very little additional computation expending. But, at the same time, it also has some problems including easy trapping into local minima and bad compression performances. In this paper, several novel search strategies used in the proposed OTA are proposed; actually, all these search strategies can be used in other algorithms. These search strategies include new adaptive termination threshold, motion vector (MV) prediction, zero MV priority and search templates. Extensive standard test sequences have shown that the proposed search strategies and algorithm is very effective, robust and adaptive. We focus on all the performances of coding image but not only on bit-rate or PSNR.

摘要： 之所以块匹配算法一直都是视频压缩处理技术中研究的焦点，是因为它的有效性和易于实现性。在过去的几十年里，人们提出了很多快速搜索算法来代替全搜索算法以达到实时视频处理的目的。其中，速度最快的应该是一步一次搜索(OTA)方法，它可以利用很少的几个点就可以得到运动矢量，而且实现起来也特别简单，额外开销很小。但它同时也存在着很多缺点：很容易陷入局部最小，图像压缩质量和码流大小都不是很理想。本文采用全新的搜索策略对其进行改进，包括利用全新的自适应搜索提前终止门限、运动矢量预测、零运动矢量优先、以及搜索模版技术。当然，这些策略完全可以运用到其他算法当中。大量的标准序列测试结果表明了这种算法的有效性、强健性和高度自适应性。本文的搜索策略和算法关注编码后的所有性能，而不仅仅是只关注编码速度或者是峰值信噪比。