Fast Motion Vector Generation for Video Coding by Gray Prediction
Yung-Gi Wu, Guofeng Huang · 2007
In this paper, we propose an efficient prediction algorithm for motion vector in video compression. Motion estimation is an important part of any video processing system. Exhaustive block matching algorithm (EBMA) can get the optimal solution; however, it takes too much computational burden. In the proposed method, we use gray prediction to get the motion vectors. Gray prediction can predict the motion vectors quickly and accurately. Several video sequences are used to evaluate the performance. Experimental results show that the time needed by the proposed method is only 1.1% compared to EBMA and 3.25% to three steps searching (TSS) algorithm while the degradation of PSNRY compared to EBMA is about 0.8 dB at most for those test sequences.