In this paper, a hybrid table look-up method for H.264 Coeff_Token Decoding is presented. In the proposed method the probabilities of the codewords with various lengths are analyzed, and based on the statistics a hybrid look-up table is constructed. In the coeff_token decoding process, firstly, a few bits are read from the bit-stream, if a matched codeword is found in the first look-up table, the further look-up steps will be skipped. Otherwise, more bits need to be read and looked up in the second table, which is built upon the number of leading 0's before the first number one. Experimental results on the RTSM Emulation Baseboard ARM926 of RealView show that the proposed method speeds up CAVLD of H.264 by about 8% with more efficient memory utilization, when compared to the prefix-based decoding method. And compared with the pattern-search method based on hashing algorithms adopted in the newest version of FFMPEG, the proposed method reduces memory space by about 77%.
[1]
Jiun-In Guo,et al.
Joint algorithm/code-level optimization of H.264 video decoder for mobile multimedia applications
,
2008,
2008 IEEE International Conference on Acoustics, Speech and Signal Processing.
[2]
Yong-Hwan Kim,et al.
Memory-efficient H.264/AVC CAVLC for fast decoding
,
2006,
IEEE Transactions on Consumer Electronics.
[3]
Zhang Xiu-Li,et al.
An improved group-based optimizing of CAVLC decoder
,
2009,
2009 International Conference on Image Analysis and Signal Processing.
[4]
Shau-Yin Tseng,et al.
A Pattern-Search Method for H.264/AVC CAVLC Decoding
,
2006,
2006 IEEE International Conference on Multimedia and Expo.