Learning Graph-Convolutional Representations for Point Cloud Denoising

TL;DR

This paper introduces a graph-convolutional neural network for point cloud denoising that effectively handles permutation invariance and outperforms existing methods, especially under high and structured noise conditions.

Contribution

It presents a novel fully-convolutional graph-based neural network that dynamically constructs neighborhoods for improved point cloud denoising.

Findings

Outperforms state-of-the-art methods on multiple metrics

Improves surface normal estimation quality

Robust against high and structured noise

Abstract

Point clouds are an increasingly relevant data type but they are often corrupted by noise. We propose a deep neural network based on graph-convolutional layers that can elegantly deal with the permutation-invariance problem encountered by learning-based point cloud processing methods. The network is fully-convolutional and can build complex hierarchies of features by dynamically constructing neighborhood graphs from similarity among the high-dimensional feature representations of the points. When coupled with a loss promoting proximity to the ideal surface, the proposed approach significantly outperforms state-of-the-art methods on a variety of metrics. In particular, it is able to improve in terms of Chamfer measure and of quality of the surface normals that can be estimated from the denoised data. We also show that it is especially robust both at high noise levels and in presence of structured noise such as the one encountered in real LiDAR scans.