A feature-based intra mode decision in H.264/AVC

This paper proposes a novel fast algorithm for reducing the computational complexity in intra mode prediction process of H.264/AVC. The proposed method makes use of the temporal correlation between adjacent frames. It is observed that most of selected modes of the co-located macroblock (MB) in the previous frame are the same as the best modes of MB in the current frame and that the sum of the absolute transformed difference (SATD) of the best mode tends to have a small value. Using these characteristics, we can guess which mode is most likely to be the best mode and this helps to reduce the number of candidate modes. Experimental results show that the proposed algorithm can reduce the computational complexity by about 78% with the degradation of the peak signal to noise ratio by less than 0.065 dB on the average compared to JM 12.2.