Precise Estimation of Probabilities in CABAC Using the Cauchy Optimization Method

Abstract

The algorithm of entropy coding that is now widely used in video compression is the Context-based Adaptive Binary Arithmetic Coding (CABAC). This paper presents a modified, improved version of CABAC, called the CABAC+. The author's idea for the improvement is to use in CABAC a more accurate estimation of probabilities of data symbols. The basis of the proposal is to use the Cauchy optimization method, in order to minimize the number of bits that are produced by the entropy encoder. The application of the CABAC improvement in the High Efficiency Video Coding (HEVC) technology increased the compression efficiency of entropy coding by 0.6% to 1.2%, depending on the parameters of the method and scenario of experiments. The use of the proposed solution increases the decoding time of a video, but the method virtually does not change the complexity of a video encoder.

Keywords

• Probability estimation
• entropy coding and arithmetic coding
• improved CABAC
• video compression