Selective mathematical modelling algorithm for quarter-pixel motion estimation

Abstract

Proposed is an error surface based selective mathematical modelling algorithm for quarter-pixel motion estimation during video encoding. In this method, first the tendency of the error surface is assessed. Using the strength of the edge at the error surface, the error surface region is classified as plain or textured. For plain regions, interpolation-free mathematical modelling is appropriate for the quarter-pixel motion estimation method. For textured regions, half-pixel interpolation is needed for more accurate modelling. Additional half-pixel SAD values are calculated and more accurate modelling is completed to find the best motion vector (MV). Experimental results show that the proposed scheme is faster than conventional modelling algorithms, and the PSNR performance is also better than that of competitive algorithms.