Scene-based low delay HEVC encoding framework based on transparent composite modeling

Abstract

To improve the rate-distortion (RD) performance of low delay (LD) High Efficiency Video Coding (HEVC) encoding while enabling scene-based non-linear editing feature (NLEF), in this paper, we present a new scene-based LD HEVC encoding framework in which scene change (SC) detection and coding are conducted jointly. A frame in a sequence to be encoded is deemed a SC if there is a sudden change in the residual energy. SCs are detected using the residual image parameters identified from the newly proposed Laplacian Transparent Composite Model (LPTCM). When a SC is detected, it is designated as an I frame; the quantization parameters for this frame and its subsequent frames before the next SC are determined. Compared to the default LD configuration, experimental results show that an average gain of 0.24 dB in RD performance and reduced overall encoding time are achieved while providing the scene-based NLEF.