Fast Region-Adaptive Hierarchical Transform with Cross-Component Prediction and Coefficient Reordering
Xiaoyu Liu, Junjie Wang, Wei Zhang · 2024
With the increasing applications of 3D point clouds in various systems, research into point cloud compression (PCC) has gained momentum in recent years. Among the point cloud attribute compression methods in the literature, Region-Adaptive Hierarchical Transform (RAHT) in Geometry-based Point Cloud Compression (G-PCC) standard, currently undergoing standardization by MPEG, stands out as a promising scheme. This paper introduces several enhancements to RAHT. Firstly, a cross-component prediction technique is proposed to reduce the coefficient redundancy by leveraging correlations between chroma color components. Secondly, to improve the entropy coding efficiency, the distribution characteristics of RAHT coefficients are analysed, leading to a method for optimizing the coding order of transform coefficients. Additionally, the RAHT coding process is accelerated through early prediction termination based on node occupancy in the RAHT tree. Experimental results show that the proposed enhancements achieve approximately 1.0% coding gains on the Luma component and 7.0% on the Chroma component, while reducing the attribute coding time by approximately 20%.