Fast HEVC inter-frame coding based on LSTM neural network technology

Abstract

High Efficiency Video Coding (HEVC) is the most commonly used video coding standard. However, its high coding complexity is a heavy burden for real-time video applications. But, coding tools designed based on traditional coding frameworks have reached limits. Furthermore, existing low-complexity video coding methods have not thoroughly analyzed the characteristics of compressed video, making it impossible to develop targeted models to reduce coding complexity. Therefore, in this research, a fast HEVC inter-frame coding technique is proposed. Firstly, we perform a characteristics analysis of HEVC inter-frame coding to explore the correlation between video frames. Secondly, we develop an Inter-Frame Feature Transfer-Long Short-Term Memory (IFFT-LSTM) model to obtain the optimal coding tree unit (CTU) partition structure. Thirdly, we embed the IFFT-LSTM model into the HEVC test platform. The experimental results show that suggested method can effectively reduce the HEVC inter-frame coding complexity with a small amount of rate-distortion (RD) performance loss while maintaining acceptable subjective video quality.