A Block-Mean Difference Based Sorting Scheme on Lossy Partail Distortion Search for Fast Optimal Motion Estimation

This work presents an efficient lossy partial distortion search (PDS) algorithm named adaptive mean difference based partial distortion search (A-MDPDS). The proposed A-MDPDS algorithm reduces computations by using a halfway-stop technique in the calculation of block distortion measure and applying a diagonal search pattern for stationary or quasi-stationary candidate blocks. For the matching point reduction, a block is divided into 4 × 4 sub-blocks that each sub-block sorted by subtracting the block mean value. Therefore, the mean difference pixels are retrieved one at time to obtain the accumulated partial SAD used as a constraint for checking the validity of a candidate block. The proposed scheme can accelerate the convergence speed and efficiently eliminate the impossible candidates earlier, resulting in substantial computation reduction. The experimental results show the proposed algorithm reduces the check pixels by about 8.4 times on average compared with the typical PDS when the motion block size is 16×16 and the search range is ±7. Compared with other lossy PDS algorithm such as NPDS, which achieved reductions of 2.4 times on average, reductions in computational complexity were achieved.