Efficient Partial Distortion Algorithms with Sorting Order of Calculation for Motion Estimation

In order to accelerate the motion estimation process, the normalized partial distortion search algorithm calculates the partial distortion by constantly selecting specific pixels from every sub-macroblock to early reject the incorrect motion vector. However, these selected pixels are supposed that the pixels' values of sub-macroblock are uniform distribution. This paper proposes a partial distortion search algorithm that joints the motion correlation of neighbored macroblocks and the sorting order of calculation with non-uniform distortion. The proposed algorithm first finds the coarse motion vector using the motion correlation of neighbored macroblocks to accelerate the motion estimation process and then calculates the partial distortion by replacing the order of calculation of normalized partial distortion search algorithm with the sorting order of calculation we proposed to early reject the impossible motion vectors. In addition to increase the probability of rejecting the impossible motion vectors without extra computations, the proposed method has a significantly lower computation and better objective quality than traditional algorithms. 