基于起点预测的快速运动估计算法

A fast motion estimation algorithm based on initial search point prediction

在研究菱形法等算法的基础上,利用序列图像的相邻块运动矢量的高度相关性和运动矢量的中心偏移特性,提出了一种基于起点预测的快速运动估计算法.该算法设计了菱形和正方形两种模板,在搜索过程中根据图像的内容(运动类型)采用组合模板进行灵活处理,提高了快速块匹配的搜索速度.该算法具有基于内容搜索的特点.实验结果表明,该算法在速度和准确性方面都优于传统的快速运动估计算法.

The motion estimation is a key technique used in video compression. By studying the diamond search algorithm, etc., this paper presents a new fast motion estimation algorithm based on initial search point prediction, called the prediction and combinative pattern algorithm (PCP) in which the high correlation of adjacently blocks motion vectors and the centerbiased characeristic of motion vectors in image sequences are used. The algorithm designs two kinds of patterns, one being diamond and the other square, and then combines the patterns to process the image according to its content (motion kind), which significantly speeds up the block matching. The algorithm has the characteristic of contentbased searching. Experimental results show that the proposed algorithm excels other traditional algorithms in speed and veracity.