Reducing Computational Complexity of Decoding-assisted HEVC Inter Prediction

• Main Authors: Cheng-En Lee, 李承恩
• Other Authors: Yinyi Lin
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/hnvfm9

碩士 === 國立中央大學 === 通訊工程學系 === 106 === In recent years, a lot of high resolution multimedia devices have been developed. In order to compress the significant data effectively, HEVC utilize multiple techniques to decrease bitrate efficiently. Review the previous research, using Sum of Absolute Bi-Prediction Differences(SABPD) and Template Matching(TM) algorithm in extended merge mode decision and inter mode decision, the Decoding-assisted HEVC utilize encoded information to predict. Decoder can predict with same information, so we can reduce transmitting header information to decrease bitrate. However, the coding time might be longer due to the increase of computational complexity. In this thesis, in order to reduce computational complexity, we proposed the Zero-block algorithm is utilized to early terminate the segmentation of Coding Unit(CU), and on the other hand, weighted RDO threshold, Coded Block Flag(CBF) and motion vector difference are utilized to early terminate the prediction mode decision in merge mode and Inter2Nx2N mode. Finally, we combine these two algorithm. According to experiment results, our proposed algorithm achieves that encoding time was reduced by 55.85% and BDBR only increased by 0.894%. Compare with HEVC, we can not only reduced encoding time by 48.58% but also decreased BDBR by 2.355%.