An efficient block matching algorithm for fast motion estimation in video compression

Motion estimation is the most time-consuming module in any video coding standard. Existing block-matching algorithms (BMA) reduce the-search time of the motion estimation process with moderate loss in quality. In this paper, a novel block-matching algorithm for fast motion estimation namely direction-based block-matching algorithm (DBM) is proposed. Existing BMA initiate its search from the center of the search area for every frame, which is time consuming. The proposed algorithm utilizes the temporal correlation in the motion vectors between successive frames to reduce the search time. The motion vectors of the previous frame imply the direction of motion in the succeeding frames thus minimizing the search domain. Experimental results show that DBM provides faster searching with moderate to low distortion compared to existing BMA.