Non-Local Cross-Component Prediction for Video Coding beyond VVC

In versatile video coding (VVC) and enhanced compression model (ECM), which is developed by JVET to explore novel techniques beyond VVC, cross-component prediction (CCP) methods derive CCP models with local samples. This paper presents a method of non-local cross-component prediction (NL-CCP), where the cross-component relationship can be estimated with non-local information. First, the CCP model can be inherited from a CCP-coded neighbouring block. Second, the CCP model of a previous CCP-coded block is stored in a history-based table, which can be fetched and used by the current block. Finally, a candidate list is built with these two kinds of NL-CCP candidates. The encoder can select the optimal candidate and send an index to the decoder. Experimental results show that NL-CCP provides an average BD-rate saving of 0.19%/2.12%/2.30% on Y/Cb/Cr components with a negligible change on the running time, compared with ECM-8.0 in all intra configurations. NL-CCP has been adopted into ECM.