A quadratic prediction based fractional-pixel motion estimation algorithm for H.264

In this study, a quadratic prediction based fractional-pixel motion estimation (ME) algorithm for H.264 is proposed. Here, a "degenerate" quadratic function is used to determine the "best" quantized predicted fractional-pixel motion vector (MV) for a variable size block. Based on the partial probability distribution of the sum of absolute component differences between the best fractional-pixel MV determined by the full search ME algorithm and the "best" quantized predicted fractional-pixel MV determined by the proposed algorithm, a small diamond search pattern (SDSP) is used to determine the final best MV at 1/4-pixel accuracy and the SDSP will be applied at most 3 times. Additionally, if both the best quantized predicted fractional-pixel MV determined by the proposed algorithm and that determined by the center biased fractional-pixel search (CBFPS) algorithm are identically (0,0), (0,0) is directly determined as the final best MV at 1/4-pixel accuracy. Based on the experimental results obtained in this study, the four ME performance measures of the proposed algorithm are better than that of four comparison algorithms, with slight degradations in average PSNR and bit rate.