Abstract: Geometric features such as creases and boundaries of mesh models are sensitive to human eyes.However,existing simplification methods tend to ignore these important geometric features.Thus,they may lead to visual distortion in the simplified model.To address the problem,we present a ridge-valley line based simplification method.First,the method extracts ridge-valley lines via implicit surface reconstruction of the original model.Then,the geometric feature points that contain important geometric information are extracted from the ridge-valley lines.Finally,a multi-scale compactly support radical basis functions（CS-RBFs）approach is adopted to reconstruct the implicit surface of the simplified model.Implementation results show that the proposed method can simplify the model efficiently and accurately,meanwhile it can reduce the number of vertices significantly compared with N-Garland method.