A quadratic Prediction Based Fractional Pixel Search Scheme for H.264 Video Coding

• Main Authors: Jing-Fu Chang, 張君甫
• Other Authors: Jin-Jang Leou
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/53884959446446612358

碩士 === 國立中正大學 === 資訊工程研究所 === 92 === The joint video term (JVT) of ISO/IEC MPEG and ITU-T VCEG develops a new standard for coding natural video images, namely, H.264, in which fractional pixel motion estimation (ME) and compensation are used to achieve more accurate motion description and higher compression efficiency. However, the complexity of fractional pixel ME will increase. At 1/4-pixel accuracy, if the full fractional pixel ME search algorithm is employed, a total number of 16 fractional pixel positions are required to be examined. But many fast integer pixel ME search algorithms have decreased the search positions to about ten. To reduce the computation complexity of fractional pixel ME is greatly desired. In this study, a fast fractional pixel ME search scheme for H.264 is proposed. In this study, a “degenerate” quadratic function is used to directly determine the “best” quantized predicted motion vector (MV) at 1/4-pixel accuracy for a variable size block. Based on the partial probability distributions of the sum of absolute component differences between the best MV determined by the conventional 2-stage full ME search algorithm for 1/4-pixel accuracy and the “best” quantized predicted MV determined by the proposed scheme, a small diamond search pattern (SDSP) at 1/4-pixel accuracy is used to determined the final result and the SDSP will be applied at most 3 times. In the proposed modified scheme, if the best quantized predicted MV determined by the proposed scheme and that determined by the center biased fractional pixel search (CBFPS) algorithm are identically (0,0), then (0,0) is directly determined as the MV at 1/4-pixel accuracy for the variable size block, without applying the SDSP. Based on the experimental results obtained in this study, the four ME performances of the proposed scheme and the proposed modified scheme are better than that of four comparison approaches, with slight degradations in average PSNR and bit rate. This shows the feasibility of the proposed scheme.