Modifying ARPS method by use of global motion information for fast and precise motion estimation

Abstract

In this article we introduce a novel motion estimation method for video sequences. Proposed method can be employed as a part of video encoders, and decoders to increase the compression ratio, and decrease the temporal redundancies. In the presented method, first a global motion vector is computed for each frame, by use of a fast procedure. This step is in fact, the key to the whole algorithm, because its result affects the accuracy and computational load of the next steps. The obtained motion vector is then directly attributed to the blocks with the global motion. To deal with the blocks whose motions are not equal to the global motion, a fast-search method, dubbed M-ARPS (M-ARPS), is suggested. Finally, by performing thorough experiments we demonstrate the supremacy of the proposed method in relation to some popular block-based motion estimation algorithms.