A block type decision algorithm for H.264/AVC intra prediction based on entropy feature
Guifen Tian, Tianruo Zhang, Xianghui Wei, Satoshi Goto · 2008
Rate distortion optimization (RDO) based spatial intra prediction is an important new feature in H.264/AVC. It encodes each MB with variable block sizes and variable coding modes. Thus, drastically increased computation complexity is involved. This paper proposes an efficient block type decision algorithm. Entropy feature of each MB is extracted before intra coding. Based on this feature, proposed algorithm can decide the best block size for each MB instead of exhaustively examining all block sizes. Sufficient experiment results show that proposed algorithm achieves about 95% decision accuracy with PSNR loss less than 0.004dB; bit-rate increase is negligible and time reduction is 22% to 33%, as target frame sizes changing from QCIF to HDTV (1080p).