A New Block Based Motion Estimation with True Region Motion Field

Abstract

In this paper a new block based motion estimation algorithm is presented. The algorithm, named as block based motion estimation with true region motion field, is actually based on the successive motion estimation from the highest confidence motion vectors to lowest one. It is utilizing fact that the image pixels corresponding to the same object projection all account similar movement. By estimating motion vector first for the blocks which incorporate at least two different spatial gradient directions (corner blocks) we got reliable estimate for the movement of the other blocks (edge and flat blocks) associated with the same object projection. After than it is only the subject of correct interpolation technique with a refining step to calculate rest of the missing motion vectors. This strategy eliminates chaotic motion vectors and gives motion field more corresponding to a true motion in image sequence, while incorporating small prediction PSNR increase against fast algorithms.