Research And Application On Feature Preserving Mesh Denoising

The development of 3D reconstruction technology has effectively reduced the acquisition cost of 3D mesh models,and promoted the application of 3D mesh models in various fields.The 3D mesh denoising is the most basic and the most important step in the 3D reconstruction process.Therefore,the research on the 3D mesh denoising is of great significance in the field of computer vision and computer graphics.In particular,how to effectively preserve the original geometric features of the model during denoising is always the focus in the research of mesh denoising.This paper classifies and introduces the existing feature detection methods,denoising methods and the application of feature preserving mesh denoising algorithm,and proposes a new feature detection method which is based on surface curvature and two feature preserving 3D mesh denoising methods: L1 optimization feature preserving denoising algorithm and feature detection based feature preserving denoising algorithm.The feature detection method based on surface curvature includes four steps: pre detection,feature selection,feature connection,and feature fine-tuning.This method can effectively detect the potential feature points of the mesh,and pick out the real feature points.The L1 optimization feature preserving denoising algorithm mainly uses the basic idea of sparse method,and it defines an energy function including three terms: position offset energy,L1 sparse feature energy and normalization energy.In this paper,the minimum value of the energy function is solved by Split Bregman method,so as to achieve the denoising of the mesh.The experimental results show that the proposed method has good denoising and feature preserving ability.The feature detection based feature preserving denoising algorithm is a promotion of the L1 optimization feature preserving denoising algorithm which combines feature detection.The greatest improvement of this algorithm is that the feature points are detected according to the current mesh in each iteration,and the existing feature points are taken into account during the vertex updating of the next iteration.The algorithm integrates feature detection into the denoising process and effectively utilizes the feature information.It's able to preserve the feature information of the mesh while denoising effectively.