Multi-View Projection Based Joint Geometry and Color Hole Repairing Method for G-PCC Trisoup Encoded Color Point Cloud

Wenxu Tao, Gangyi Jiang, Mei Yu, Yun Zhang, Zhidi Jiang, Yo-Sung Ho

Abstract: Efficient compression of color point clouds (CPC) is crucial for immersive visual systems. The G-PCC standard by Moving Picture Experts Group offers effective compression solutions for static CPCs. G-PCC Trisoup, one of the G-PCC solutions, achieves high coding performance using a pruned octree, combined with triangle soup (trisoup) surface encoding. However, encoding errors introduce distortions like triangular holes, degrading visual quality. This article proposes a multi-view projection based joint geometry and color hole repairing method for G-PCC trisoup encoded color point cloud, significantly improving reconstruction quality. Firstly, a G-PCC Trisoup encoded CPC is projected onto multi-view color and geometry maps. A multi-view projection-based scheme detects triangular holes by identifying geometry discontinuities. A geometric weighting factor, considering direction, distance, and level set distance, repairs geometry maps. Additionally, a novel gated hole repairing network repairs color maps by leveraging semantic information and reducing background interference. Finally, the joint geometry and color repaired results are inversely projected to obtain the final CPCs. The proposed method is validated on distorted CPCs encoded with G-PCC Trisoup at different levels. Experimental results demonstrate effective improvement in geometry and color objective quality, leading to enhanced visual perception.