End-to-End Distortion Modeling for Error-Resilient Screen Content Video Coding

Abstract

To improve the compression performance of screen content coding, extension coding standards (HEVC-SCC, VVC-SCC) have been developed. However, considering the compression ratio alone may lead to packet losses in bitstreams which may cause plenty of images decoded incorrectly, degrading the video quality at the receiver side. Thus, it urgently needs to study source-channel jointly coding scheme of screen content video. The most significant challenge lies in the complex spatial-temporal characteristics of screen content video, which complicate the creation of an accurate end-to-end distortion model. In this article, we delve into the traits of screen content video and construct an end-to-end distortion model. Building upon this, we introduce an error resilient coding scheme specifically for screen content video. More specifically, we first consider the characteristic of non-stationary temporal domain variation and classify the screen content images into three types of frames using a fast block-searching method. We then propose an adaptive error concealment method, taking into account the spatial-temporal prediction characteristics. Following this, we derive a pixel-level end-to-end distortion model and incorporate it into the rate distortion optimization process. Our experimental results reveal that, compared to state-of-the-art methods, our proposed method significantly enhances both objective and subjective quality across a variety of channel conditions.