A matching criterion for motion compensation in the temporal coding of video signal

Abstract

Video data have spatial as well as temporal redundancy and motion estimation plays a vital role in the removal of temporal redundancy of video data. Block matching techniques are mostly used and generally the matching criterion in these block matching techniques is the mean absolute error (MAE). Though MAE-based approach is simple and less complex, it does not give better prediction specially for large motion video inputs with contrast variation. In this manuscript, a new block matching criterion has been suggested and experimentally compared with three existing methods in terms of four parameters—average MAE/pixel, average search points/block, average peak signal to noise ratio (PSNR) and average number of bits/pixel value. Proposed criterion gives nearly 75% less average error than conventional MAE. An increase of nearly 16% in average PSNR value and 37% reduction in average bits/pixel value in comparison to MAE has been observed for the proposed criterion. Further, these criterions have also been evaluated for quality/compression ratio which is nearly 80% more for the proposed criterion than corresponding MAE metric.