Luma-Attention-based Chroma Intra Prediction for Versatile Video Coding
Bumyoon Kim, Yong‐Seong Kim, Hyunki Jeong, Byeungwoo Jeon · 2025
The remarkable advancements in machine learning are impacting most areas of research, and video coding is no exception. In this paper, we study a neural network-based chroma intra prediction technique which utilizes weighted coefficients obtained only from luma attention information. Experimental results show coding gains of 0.45%, 1.46%, and 1.28% for the Y, Cb, and Cr channels compared to VVC Test Model (VTM) version 23.0. These results highlight potential of chroma prediction solely from luma information as a novel approach to chroma intra prediction.