Periodic Enhanced Frame Based Long-Short-Term Reference in HEVC for Conference and Surveillance Video Coding

Abstract

The coding tools of the High Efficiency Video Coding (HEVC) standard were essentially designed for ordinary video contents. They have not fully exploited the long-term temporal dependence of surveillance videos and conference videos, in which the backgrounds are generally stationary. In this paper, a periodic-enhanced frame-based long-short-term reference scheme is presented for coding the videos with stationary backgrounds. First, periodic-enhanced frames are proposed based on global rate-distortion optimization, in which the background error propagation characteristic is introduced to reduce the long-term temporal redundancy. Second, two limitations, such as temporary occlusion and quantization error, are analyzed as applying the enhanced frame on a default HEVC hierarchical prediction structure. In order to reduce the limitations, a long-short-term reference scheme is presented based on the enhanced frames. In this scheme, the periodic-enhanced frames are employed as a long-term reference picture, and the adjacent frames are set as short-term reference pictures. The proposed method is verified by testing on the conference videos and surveillance videos.