Using CNN for Encoder Optimization in H.265/HEVC

Abstract

In this work-in-progress paper, we proposed using deep learning techniques, especially the deep Convolutional Neural Network (CNN) to perform critical tasks of video ending within the framework of H.265/HEVC. Deep CNNs have achieved breakthrough improvements on image recognition tasks such as image classifications, object identifications, and image annotations. However, very few work has been done in applying deep CNN to video encoding. In order to take advantage of the significant capabilities of deep CNN on image content detection, we proposed using deep CNN as the primary technique to perform critical tasks in video encoding that are relevant to the contents of one or multiple video frames. More specifically, we designed deep CNNs for the following tasks in H.265/HEVC encoder: partitioning CTU to CUs; partitioning CU to PUs; performing intra prediction; and performing inter predictions.