A Study on Improving Luma-based Chroma Intra Prediction Method

Abstract

In luma-based chroma intra prediction in an HEVC test model, chroma samples are predicted from down-sampled luma samples of the same block by a linear model relationship. We propose improving the prediction performance by applying horizontal filtering to luma samples which removes the horizontal aliasing effect. Experimental results using HM-3.0 had up to a 1.7% BD rate improvement of the chroma signals.