A high throughput binary arithmetic coding engine for h.264/avc

Abstract

Context-based adaptive binary arithmetic coding (CABAC) adopted by the latest video coding standard- H.264/AVC main profile, can achieve good compression performance both in encoding and decoding H.264 syntax element. However, the binary coding engine of CABAC is bitwise in H.264/AVC, which leads to the limited throughput of CABAC encoder. Also the traditional iterative scheme for renormalization in binary arithmetic coding can result in degradation of coding efficiency. To avoid more costly operation in CABAC, an efficient coding design for handling high-resolution video sequence is strongly required. This paper proposes an optimized parallel processing approach and high throughput architecture design for renormalization in binary arithmetic coding engine, which is well suited for efficient software and hardware implementation over H.264/AVC HDTV encoder. Comparing with the traditional design, the results show that this novel approach makes a symbol encoded in one cycle; and for a set of test sequences, the throughput rate has been increased nearly 20% on average