Adaptive Cross Component Linear Model for Chroma Intra-Prediction in VVC.

The Cross-Component Linear Model (CCLM) is an intra prediction technique that is adopted into the latest Versatile Video Coding (VVC) standard. CCLM assumes a linear correlation between the luma and chroma components in a coding block. With this assumption, the chroma components can be predicted by the linear model (LM) mode, which utilizes the reconstructed samples in the luma channel as well as neighboring samples of the chroma coding block. In this paper, new two methods are proposed for CCLM. First, we introduce an improved CCLM approach based on univariate polynomial regression, which is suitable for dealing with the situation where the distribution of sample points in the coding unit is relatively discrete, to achieve a more accurate prediction of chroma components. Second, to reduce the computation complexity of chroma intra prediction, a fast chroma intra prediction mode optimization algorithm based on CU texture complexity is proposed, which aims to reduce the number of candidate mode types for chroma intra prediction in advance by calculating CU texture complexity. Experimental results show that the proposed method can achieve the BD-rate saving of 0.01%, 0.34%, and 0.41% on average for Y, Cb , and Cr components with All Intra (AI) configuration. At the same time, the average time saving of this method reached 6.71% compared with the baseline algorithm.