Coarse-to-fine partial distortion search algorithm for motion estimation

This study proposes a coarse-to-fine partial distortion search algorithm for motion estimation to accurately predict the minimum distortion region. In the proposed method, the distortion behavior is first analyzed in a search window to indicate the possible occurrence region of the global minimum distortion position. Normalized partial distortion searching is then performed on the selected region, instead of the whole search window. The proposed algorithms have a low computational complexity yet do not degrade the PSNR significantly.