Mesh segmentation by combining mesh saliency with spectral clustering

In this paper, we present a new mesh segmentation method that achieves visually meaningful segmentation by combining mesh saliency with spectral clustering. Our method solves the segmentation problem by embedding the original mesh model into spectral space. Firstly, the mesh concave regions are determined according to the minimum rule in visual theory, and then a Laplacian matrix is defined by considering the mesh saliency and curvature information. Next, we calculate the first k eigenvectors of the Laplacian matrix by eigen-decomposition process, and embed the original mesh into a k-dimensional spectral space. Finally, we can achieve the visually meaningful segmentation by utilizing the Gaussian Mixture method, and the initial cluster centers are decided by mesh saliency. The experimental results have demonstrated the effectiveness of the proposed segmentation method. Especially for the model with convex regions and branch components, our method can achieve better visual quality.