3D Point Cloud Denoising Based on Color Attribute

Collecting or transmitting point cloud data is often subject to noise, which can potentially affect the accuracy of geometry and color representation in different spatial domains. This study addresses the denoising problem specifically for 3D point cloud color data and proposes two dedicated denoising algorithms based on the characteristics of noise in different spatial domains. In the denoising process, the RGB color space is first transformed into the YUV color space for further denoising operations. Surface smoothing is achieved by employing either a median filter or a bilateral filter based on the impact of noise on spatial information. These algorithms, built upon 2D image processing techniques, offer two key contributions: 1) color correction on spatial points to enhance denoising performance, and 2) the use of low-complexity filters while maintaining comparable filtering effectiveness, resulting in nearly a twofold reduction in processing time.