High-speed rate estimation based on parallel processing for H.264/AVC CABAC encoder

The advanced video codec of H.264/AVC has been extensively employed for data compression for efficient video storage or transmissions. However, its intensive computational complexity causes a bottleneck for its real-time applications. Simultaneously, the sequential operation in the normal encoding process is not favorable with regard to hardware implementations. In this paper, a high-speed and accurate bit-rate estimation scheme is proposed to skip the actual entropy engine of context-adaptive binary arithmetic coding (CABAC), which is useful for reducing the computational complexity and saving the encoding time in H.264/AVC rate distortion (RD) optimization for mode decision. The characteristics of syntax elements and the principle of CABAC are firstly analyzed. Then guided by the spirit of parallel processing, a novel algorithm is constructed to estimate the bit-rate in RD mode decision. The proposed method significantly accelerates the encoding process while well maintaining the video quality, and is well suited to hardware implementations due to its feature of parallel processing¹.