Design of Block Matching Criterion and Motion Estimation Algorithm for Video Encoder

Abstract

Video coding standards have attracted much attention world wide recently with an in-creasing number of very large scale integratipn (VLSI) and software implementation of these standards becoming commercially available. In video encoders, motion estima-tion is used to remove interpixel redundancy between frames and block based motion estimation algorithms are mostly used in literature for their simplicity. In such algo-rithms, video frames are divided into fixed/variable size blocks and each block in the current frame is searched for a best match block in the reference frame in a given search window. Various types of block matching criterions like mean square error (MSE), mean absolute error (MAE), normalized cross correlation function (NCF), etc. have been discussed in literature, but MAE is being used in most of the video coding stan-dards due to its low computation and simplicity. In large motion video inputs with contrast variations, it has been found that MAE does not produce good results. Further, due to the commercialization of real time multimedia based applications like video on demand, video conferencing etc., reduction in computation cost of motion estimation has become critical requirement. Therefore, in last few years, it has been a challenge for researchers to design accurate block matching criterion and efficient motion estimation algorithm for encoding of video signals. newlineIn this thesis, a robust pixel mapping based block matching criterion has been pro-posed and its performance has been compared experimentally with three other existing block matching criterions: MAE and its variants vector matching criterion (VMC) and smooth constraint mean absolute error (SC-MAE) in terms of various performance parameters like average error/pixel, average search points/block, average peak signal to noise ratio (PSNR) per frame, average bits/pixel value and execution time.A reduction of nearly 75% in average error/pixel and an increase of 16% in PSNR (dB) have been observed for proposed criterion in comparison...