Increasing data throughput in PIPE coding using extended v2v-codes

Entropy coding schemes like context-based adaptive binary arithmetic coding (CABAC) as used in the H.264/AVC video coding standard are established in the market. To increase throughput and possibilities for parallelization, probability interval partitioning entropy coding (PIPE) was developed. It replaces the arithmetic entropy coding step of the CABAC scheme with a set of bin coders, each operating at a fixed probability. In this paper, the use of variable-to-variable length (v2v) codes for the bin coders is discussed and a concept for increasing the throughput of table-based v2v-coder implementations is developed. It is based on code word concatenation and thus virtually neutral in terms of coding efficiency. Using this technique in the emerging HEVC video coding standard show an overall decoding time reduction of up to 5% when compared to the unchanged CABAC.