Point Cloud Denoising Algorithm Based on Noise Classification

Abstract

Aiming at the problem of different kinds of noise in 3D point cloud data, we propose a point cloud denoising method based on noise classification. This algorithm first divides noise in point cloud data into inner points and outer points, and uses radius filtering and statistical filtering to remove the outer points. Then, normal and curvature information of point cloud are estimated by the principal component analysis. At the same time, curvature information is introduced into the bilateral filtering factor to improve the existing algorithm. Ultimately, we smooth inner points mixed in point cloud by utilizing the modified algorithm. Comparing the improved algorithm with the bilateral filtering algorithm on the bunny, horse and dragon model, experimental results indicate that the maximum error and the average error are reduced. The algorithm in this paper makes models' features maintained better while models are smoothed.