A Fast Algorithm for CU Depth Decision Based on the Minimum Risk Bayesian Criterion

Abstract

In the high-efficiency video coding standard, the division process of coding units is an optimal depth search process. Due to the block method of the quadtree, the depth selection process of the coding unit will consume a lot of coding time. This paper proposes an algorithm for fast selection of coding unit depth based on the minimum risk Bayesian criterion. It uses huge database information to learn Bayesian threshold and Bayesian conditional probability density offline, establishes a look-up table, and selects the most Excellent subset of coding features. This algorithm can effectively reduce the time for the current coding unit to select depth. The experimental results show that compared with the standard test software HM16.16, this algorithm saves 41.8% of the total coding time on average.