A Novel Fast Motion Estimation Algorithm Based on Starting Search Point Prediction

Motion estimation is the most time consuming part of the video coding because of its complexity. A novel fast block matching motion estimation algorithm named Long-Rood fast motion estimation algorithm based on starting search point prediction is proposed for real-time video coding. A number of standard test sequences are adopted for performance comparison of the proposed algorithm and full search (FS), three-step search (TSS), diamond search (DS) and adaptive Rood pattern search (ARPS) algorithms.The experimental results show that the proposed algorithm is about 3.29~17.24 times faster than that of TSS, 1.71~9.05 times faster than that of DS, and 1.24~4.45 times faster than ARPS, while the peak signal-to-noise ratio (PSNR) is almost the same or a little higher than the other algorithms.