Abstract: Most of the existing mesh simplification algorithms ignore some small features of the original model reconstructed by industrial CT serial images and the triangular meshes quality of the simplified model is poor. To solve this problem, a simplification method with feature preserving for mesh model was proposed. In this method, the original model was simplified by triangle collapse, and then after the average dihedral angle error of mesh model reached the allowable error, the model was simplified by edge collapse. For triangle collapse, the folding point of one triangle was determined by its normal vector, the Gaussian curvature of each vertex and vertices projection method. The collapsing cost of one triangle was determined by the sum of dimensionless number of its local volume error and its dihedral angle error. For edge collapse, the dihedral angle error was introduced in the edge collapse cost of QEM. The experiment result shows that the presented simplification method could generate the feature preserving, high quality and lower geometric error simplified model compared with most of the existing simplification algorithms.