Feature-preserving mesh denoising based on vertices classification

In this paper, we present an effective surface denoising method for noisy surfaces. The two key steps in this method involve feature vertex classification and an iterative, two-step denoising method depending on two feature weighting functions. The classification for feature vertices is based on volume integral invariant. With the super nature of this integral invariant, the features of vertices can be fixed with less influence of noise, and different denoising degrees can be applied to different parts of the pending surface. Compared with other methods, our approach produces better results in feature-preserving.

Figure optionsDownload high-quality image (67 K)Download as PowerPoint slideResearch highlights
► We introduce the integral invariant and k-means clustering for mesh denoising.
► The method is effective in both non-severe and severe noise situations.
► The algorithm gives different denoising intensities for feature and non-feature vertices.
► The disadvantages of area-weighted algorithms are overcome by our algorithm.