CNN Oriented Fast CU Partition Decision and PU Mode Decision for HEVC Intra Encoding

摘要

In the process of intra prediction of the High Efficiency Video Coding(HEVC), the calculation requirement is hugely increased. In order to solve this problem, we propose two algorithms, including a novel fast CU partition decision algorithm and a novel fast PU mode decision algorithm, for HEVC intra coding based on convolutional neural network(CNN). We can directly predict the CU partition structure without the reference to the information of its neighboring CU, and replace the RMD process with our CNN to get the PU candidate list. The experiments show that compared to the former CNN algorithms, our algorithms could save more than 20% coding time while ensuring the coding quality of pictures.