FCBHS: a fast center-biased hybrid search algorithm for fast block motion estimation

Abstract

Describes a fast block-matching algorithm (BMA) for motion estimation exploiting the center-biased motion vector distribution characteristic more efficiently than the center-biased hybrid search (CBHS) algorithm does. This proposed algorithm, which is called the "fast center-biased hybrid search" (FCBHS), employs a hybrid method of a compact plus-shaped search, an X-shaped search and a diamond search to reduce the search point for motion vectors which are distributed within a 2-pixel radius of the center of the search window. Experimental results show that the FCBHS algorithm is about 44 times faster than a full search and 7-10% faster than the CBHS algorithm. Such a fast sub-optimal motion estimation algorithm is essential for situations which have limited processing power and constraints on the amount of power consumption, like real-time mobile video coding applications. This paper compares the popular sub-optimal block-matching technique with FCBHS, for which both the processing speed and the accuracy of motion compensation are tested over widely-used H.263 test video sequences.