Low-computation and high-performance adaptive full search block-matching motion estimation

Abstract

This paper describes a low-computation and high-performance adaptive full search block matching (FSBM) by exploiting the spatio-temporal motion correlation between the video frames. In this work, previous frame block motion vector is the reference vector to track the motion of the current block. We also investigate the advantages of partial distortion measures for early jump outs to reduce the computations further. The experimental results reveal that the proposed adaptive search area FSBM algorithm (ASFSBM) saves up to 96% of computations, as compared to FSBM and the estimation accuracy is very close to that of FSBM. To obtain fine motion vector accuracy, ASFSBM has been implemented with half-pixel accuracy. Its performance has been compared with three-step search (TSS) and FSBM for a variety of test video sequences.