A fast block-matching motion estimation algorithm based on spatial-temporal motion vector correlation

Abstract

Based on the study of motion vector distribution and motion continuousness from several commonly used test video sequences, a new fast block-matching motion estimation algorithm based on the diamond search and motion vector prediction is proposed. Simulation results demonstrate that the proposed algorithm dramatically outperforms the recently published diamond search algorithm (DS) and the four-step search algorithm (4SS). Compared with DS and 4SS, the proposed algorithm requires less computation by up to move than 20% on average, while also achieving close performance in terms of mean square error.