J-SGFT: Joint Spatial and Graph Fourier Domain Learning for Point Cloud Attribute Deblocking

TL;DR

This paper presents a novel multi-scale postprocessing framework that effectively reduces blocky artifacts in reconstructed point clouds by fusing graph-Fourier representations with sparse convolutions and attention mechanisms, significantly improving visual quality.

Contribution

It introduces a new joint spatial and graph Fourier domain learning method for point cloud attribute deblocking, outperforming existing compression baselines.

Findings

Achieves 18.81% BD-rate reduction in Y channel.

Improves visual fidelity with minimal overhead.

Outperforms GPCC TMC13v14 baseline.

Abstract

Point clouds (PC) are essential for AR/VR and autonomous driving but challenge compression schemes with their size, irregular sampling, and sparsity. MPEG's Geometry-based Point Cloud Compression (GPCC) methods successfully reduce bitrate; however, they introduce significant blocky artifacts in the reconstructed point cloud. We introduce a novel multi-scale postprocessing framework that fuses graph-Fourier latent attribute representations with sparse convolutions and channel-wise attention to efficiently deblock reconstructed point clouds. Against the GPCC TMC13v14 baseline, our approach achieves BD-rate reduction of 18.81\% in the Y channel and 18.14\% in the joint YUV on the 8iVFBv2 dataset, delivering markedly improved visual fidelity with minimal overhead.