Fast mode decision based on RDO for AVS high definition video encoder
PCM'10 Proceedings of the Advances in multimedia information processing, and 11th Pacific Rim conference on Multimedia: Part II
Hi-index | 0.43 |
In A VS-P2 video compression standard, similar to MPEG-2, entropy coding firstly assembles two dimensional coefficients of each block into a sequence of (Run, Level) combinations serially. As we know, such the serial run-length method is usually undesirable for hardware accelerator and thus, this paper proposes an efficient parallel algorithm to Run-Length Coding, which can determine the (Run, Level) combinations for one row of coefficients from a block in one clock cycle. In addition, Level-based multiple VLC tables switch mechanism (Context-based VLC) is further introduced in AVS-P2 entropy coding module to identify the big variation of probability distribution of (Run, Level) combinations. As a result, table selection for coding the current Level necessarily depends on the previously coded coefficients. Thus, we propose a parallel Looking-Up Table method, which can select the tables for one row of coefficients from a block in one clock cycle. On the other hand, at RDO stage, the calculation of rate term only needs to get the number of bits for each coded signal without the knowledge of its concrete value. Consequently, in hardware design, the Looking-Up Table in pre-coding can be mapped into a series of logic operations and thus much hardware memory can be saved. At the actual entropy coding, we only need to replace the logic operation of pre-coding with the actual 2D-VLC tables. Using our proposed hardware accelerator of AVS entropy coder, the results of simulation and synthesis demonstrate that the computing complexity and memory requirements are both reduced .