RD-cost as statistical inference for early intra mode decision in HEVC

Abstract

Fast intra mode decision strategy is proposed to overcome the brute force mode decision for the coding unit (CU) in High Efficiency Video Coding (HEVC). In this work, firstly, the intra mode in HEVC is studied and various facts are present in the form of statistical evidence that will be of great importance for the future of intra mode. Secondly, a novel rate distortion (RD) cost prediction model is developed for early termination that is based on the RD cost variation in the neighboring CUs with-respect-to their co-located CUs. Finally, an early termination of intra mode decision process is predicted using this predicted RD cost. Experimental results demonstrate that the predicted RD cost has a high correlation (R2) of 96% with the original RD cost, and it is the highest accuracy achieved as compared to the existing state-of-the-art RD cost prediction methods. The proposed work saves 21% to 31% of total encoding time of the codec with Bjontegaard delta bit rate (BDBR) of 1.0% to 2.1% on average, respectively.