Fast Saliency-based Partition Decision for HEVC Coding Unit of Enlarged Coding Block Sizes

• Main Authors: Shih, Chia-Hao, 石家豪
• Other Authors: Wei, Ching-Huang
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/g5y28g

碩士 === 國立高雄第一科技大學 === 電腦與通訊工程系碩士班 === 106 === To obtain the best partition of coding tree unit (CTU) structure with the lowest rate-distortion (RD) cost in the HEVC video compression standard, the computational complexity is very high for the computation of RD costs of all prediction modes and the comparison among these RD cost needed by each coding unit (CU). Therefore, in this paper an algorithm is proposed to reduce the computational complexity of HEVC coding using early skipping larger-sized coding unit, human visual saliency patterns instead of the RD cost, Gradient information statistics 33 intra-angular prediction modes, and the coding block size enlarged to be suitable for the 4K ultra high-definition (UHD) video coding, so as to reduce the computational complexity and save the coding time. The experimental results show that the proposed algorithm can save 47.68% of total encoding time on average and reduce 2.43% of bit rate on average compared to iii the original mode decision in the HEVC reference software HM16.9, with the penalty of 0.163dB of the quality loss of reconstructed video on average.