An Efficient Feature-preserving Mesh Denoising Method

Mesh denoising,which aims at acquiring clean and high-quality outputs from the input noisy meshes.Mesh denoising has gained increasingly attention based on the fact that the models obtained by the scanner is inevitably polluted to some extent,which will bring unpredictable bad consequences for subsequent mesh using.In recent years,researchers have proposed many mesh denoising methods which can produce encouraging results.However,these mesh denoising algorithm had limitations in different extent,there were two main points:(1)large noise models could not keep geometric features;(2)although the algorithm achieved ideal effect,but the computational efficiency was very low.In this paper,a highly efficient approach for mesh denoising while preserving geometric features is proposed.The method of this paper is divided into four steps:initial vertex filtering,face normal estimation,vertex normal estimation and vertex update.At the initial vertex filtering step,a fast iterative algorithm is proposed to substantially reduce noise interference.According to the mesh model from the above step,use an unstandardized bilateral filter to efficiently smooth the mesh model's face normals.In the next step,classify the vertex according to the filtered face normals,so as to estimating vertex normals more accurately.Finally,update vertex by utilizing both the smoothed face normals and vertex normals obtained from above steps,the vertex positions were updated by a novel iterative algorithm.The qualitative and quantitative results show that the method proposed in this paper outperform selected state of the art algorithms,especially in the computational efficiency.