A String Prediction-Based Algorithm to AVS3 Screen Content Coding

Abstract

String Prediction (SP) is a special intra prediction mode for screen content coding, which can provide better coding efficiency while keeping lower coding complexity. In this paper, we present a string prediction-based improved algorithm to third-generation of the Audio Video Coding Standard (AVS3) screen content coding (SCC). The main content is to use the string vector (SV) of history-based block vector prediction (HBVP) as string vector prediction (SVP), and obtain the best SV by comparing the rate-distortion (RD) cost, which comes from CU-level estimation search and pixel-level motion estimation search. SV is equal to the sum of SVP and string vector difference (SVD). While the encoder only needs to signal the SVD and the HBVP index to the decoder. The experiments show that with little coding complexity increases, the algorithm presented in this paper achieves an average Y BD-rate of -5.02% (TGM)/ -3.06% (MC), U BD-rate of -5.10% (TGM)/ -3.39% (MC), V BD-rate of -5.08% (TGM)/ -3.66% (MC) from AVS SCC common test condition (CTC) test suite in all intra configuration.