Research Of Feature-preserving Mesh Denoising Algorithm

Mesh denoising,which aims at achieving high-quality 3D models from meshes corrupted with noise,is one of fundamental problems in computer graphics.With the widespread of 3D scanning devices,3d models can be captured by a variety of ways.The emergence of noise is unavoidable in those data captured from scanning real physical models.The sources of noise are numerous including measuring,algorithmic errors and limited sampling resolution.Therefore,an effective mesh denoising algorithm is indispensable for further mesh processing.It is important for mesh denoising to maintain the geometric features of the surfaces,and ensure the recovered mesh will not shrink when removing noises.Recently,researches have made significant progress in mesh denoising,but denoising still remains challenging due to its complexity in several aspects.First,in the presence of noise,how can we identify feature and non-feature vertices? Second,how should we effectively perform noise removal for feature vertices while preventing feature blurring? Third,how do we remove the sparse strong noise points on flat region.Fourth,the selection of the optimal parameter of mesh denoising algorithm is difficult.In this dissertation,we present effective mesh denoising methods for the above problems.For the first three questions,mesh denoising based on collaborative filters are proposed whose effectiveness stems from several aspects: 1)Global preprocessing is performed to drastically reduce noise influences.2)Normal tensor voting is utilized to classify the vertices,so that we can select different filters to estimate the face normals according to the detected type of vertices.Our method adaptively prevents the side effects from facets with high geometrical disparity in the feature region,which avoids the subjective selection of parameter values to achieve the local optimum,and removes outliers in the non-feature region.3)Normal difference weights are introduced to vertex updating.For the last question,dynamic and adaptive scheme for mesh denoising is proposed which automatically select the optimal parameters for normal estimation according to the local variance of normal vectors of mesh.Benefited from the well-designed filters on different types of vertices,mesh denoising algorithm based on collaborative filters produces visually and numerically better denoising results than the existing typical approaches for both CAD and generic models corrupted by high level of noises,especially at sharp features,such as edges and corners.The proposed dynamic and adaptive denoising scheme also achieves good results when dealing with all kinds of meshes corrupted by various degrees of noises.