Progressive Point Cloud Denoising with Cross-Stage Cross-Coder Adaptive Edge Graph Convolution Network

Wu Chen, Hehe Fan, Qiuping Jiang, Chao Huang, Yi Yang

Abstract: Due to the limitation of collection device and unstable scanning process, point cloud data is usually noisy. Those noise deforms the underlying structures of point clouds and inevitably affects downstream tasks such as rendering, reconstruction and analysis. In this paper, we propose a Cross-stage Cross-coder Adaptive Edge Graph Convolution Network (C$^{2}$AENet) to denoise point clouds. Our network uses multiple stages to progressively and iteratively denoise points. To improve the effectiveness, we add connections between two stages and between the encoder and decoder, leading to the cross-stage cross-coder architecture. Additionally, existing graph-based point cloud learning methods tend to capture local structure. They typically construct a semantic graph based on semantic distance, which may ignore Euclidean neighbors and lead to insufficient geometry perception. Therefore, we introduce a geometric graph and adaptively calculate edge attention based on the local and global structural information of the points. This results in a novel graph convolution module that allows the network to capture richer contextual information and focus on more important parts. Extensive experiments demonstrate that the proposed method is competitive compared with other state-of-the-art methods. The code will be made publicly available.

Primary Subject Area: [Experience] Interactions and Quality of Experience

Relevance To Conference: As 3D sensing technology advances swiftly, point clouds have been widely applied in the multimedia field. Due to the limitation of collection device and unstable scanning process, point cloud data is usually noisy. This noise presents substantial challenges for subsequent tasks in the multimedia field. In this paper, we propose a Cross-stage Cross-coder Adaptive Edge Graph Convolution Network (C$^{2}$AENet) to denoise point clouds. Our method effectively promotes efficient information flow across different denoising stages while enhancing the performance of graph convolution. Extensive experiments demonstrate that the proposed method is competitive in metrics compared with other state-of-the-art methods.

Submission Number: 1504