Abstract: Based on the research on the Mean Normal Filter for the mesh denoising, an improved hybrid denoising algorithm is proposed. Firstly, the vertices are classified into the noisy vertex and regular vertex using the arithmetic mean of the dihedral angle associated with the vertex’s all regular edge. Then, the displacement vector of the noisy vertex is that of the Laplacian method, and the displacement vector of the regular vertex is that of the Mean Normal Filter. Experimental results demonstrate that this hybrid method can efficiently suppress small and big noises while preserving the feature of the mesh during the denoising process.

Key words: computer graphics, Laplacian mesh denoising, mean normal filter, feature-preserved