Bit reduction based matching criterion for motion compensation in video coding

Abstract

Video data compression uses interframe coding to remove temporal redundancy and motion compensation plays a vital role in it. Generally, the block matching techniques are used for motion compensation and the matching criterions used are Mean Absolute Difference (MAD) and Mean Square Error (MSE). MSE has more computation than MAD and therefore, MAD is the most preferred block matching criterion for its low computation and simplicity. It has been found that MAD does not satisfies the low bit rate requirement for coding purposes as it considers the average value of the residue frame. In this paper, an efficient block matching criterion for motion compensation is being proposed and compared with MAD and MSE. Experimental results show that the proposed criterion gives better results in terms of Peak Signal to Noise Ratio (PSNR) and average bits/pixel value.