Fractal-based motion estimation for image sequence coding

In this investigation, motion estimation is carried out on three image sequences using a block matching approach. Each frame of the image sequence is partitioned into a number of fixed size blocks, and for each block the fractal dimension is calculated. For each block in the current frame, the best-matching block in the previous frame is identified using a novel two- pass searching scheme. In the first pass, the fractal dimension is calculated in nine positions within the search space. The coarse position of the corresponding block is identified based on the similarity of the fractal dimension. In the second pass, a grey level exhaustive search around the coarse position is used to determine the exact position of the corresponding block. The searching process is waived if the block has negligible movement. Preliminary results show that the new motion estimation method requires much less computation than the exhaustive search technique and provides a better estimate than the three-step search method, especially for large search spaces.