A Three-Dimensional Diffusion Filtering Model Establishment and Analysis for Point Cloud Intensity Noise

Abstract

To improve the quality of point cloud data, as well as maintain edge and detail information in the course of filtering intensity data, a three-dimensional (3D) diffusion filtering equation based on the general principle of diffusion filtering is established in this paper. Moreover, we derive theoretical formulas for the scale parameter and maximum iteration number and achieve self-adaptive denoising, fine control of the point cloud filtering, and accurate prediction of the diffusion convergence. Through experiments with three types of typical point cloud intensity data, the theoretical formulas for the scale parameter and iteration number are verified. Comparative experiments with point cloud data of different types show that the 3D diffusion filtering method has significant denoising and edge-preserving abilities. Compared with the traditional median filtering algorithm, the signal-to-noise ratio (SNR) of the point cloud after filtering is increased by an average of 10% and above, with a maximum value of 40% and above.