Pseudo-2D-matching based dual-coder architecture for screen contents coding

Abstract

Computer screens contain both discontinuous-tone content and continuous-tone content, which have essentially different characteristics. Thus, dual-coder approach is needed to effectively compress screen contents. This paper proposes a Pseudo-2D-matching (P2M) coder based dual-coder architecture for screen contents. An input Largest Coding Unit (LCU) is pre-coded simultaneously by a P2M coder and a traditional hybrid coder. The coder having best rate-distortion is selected as the final coder for the LCU. The P2M coder treats an LCU as many horizontally or vertically scanned lines and searches matching line segments in the searching-window consisting of previously coded pixels. Since the searching-window size can up to a few megabytes, to increase the chance of good matching, a hash-table is built to accelerate the matching search. The P2M coder has two matching modes: vertically scanned string matching mode and horizontally scanned 2D-shape-preserved matching mode. In the P2M coder, an LCU is precoded by two matching modes and the mode having best rate-distortion is selected. The P2M based coding architecture has significant objective rate-distortion and subjective quality improvement over traditional hybrid coding for screen contents.