2019-04-20 10:30 - 11:15
This talk presents one learning-based scheme for selecting optimal rate-control parameters in video encoding. Firstly, we will propose an automatic labeling method for optimal rate-control parameters. Then, we will propose the end-to-end deep learning method for learning the optimal rate control for each video shot. Experiments will show our scheme can save bitrates remarkably compared with the one using fixed rate-control parameters.
Video Codec is the field that relatively mature but rapid development constantly, new standard of video codec is endless, and new technology application is also changing quickly, the result of these factors is to inject young blood and energy into traditional field constantly, and it means that the engineers and researchers from the field need to learn new technology and develop new algorithm for meeting updated standard.2019 is important year for video codec, VVC standard is expected to be completed before long, more and more companies will invest in VVC that is being productized，the optimization and productization of AV1 and AVS3 will bring new scenery for industry. In addition, the devices with HDR screen and powerful GPU will become popular, and will also promote the popularization of a set of related technology and application.