Noise Intensity-Based Denoising of Point-Sampled Geometry

Abstract

A denoising algorithm for point-sampled geometry is proposed based on noise intensity. The noise intensity of each point on point-sampled geometry (PSG) is first measured by using a combined criterion. Based on mean shift clustering, the PSG is then clustered in terms of the local geometry-features similarity. According to the cluster to which a sample point belongs, a moving least squares surface is constructed, and in combination with noise intensity, the PSG is finally denoised. Some experimental results demonstrate that the algorithm is robust, and can denoise the noise efficiently while preserving the surface features.