A subdivision surface reconstruction method based on variational framework feature perception belongs to the digital geometry processing and the surface modeling field, and comprises the steps of firstly, according to the features of a model, such as the sharp angles, the sharp edges, the semi-sharp edges, etc., setting a sharpness for each edge of a control mesh, and providing a subdivision surface new rule of the feature perception; then according to the new rule, establishing a subdivision model of the subdivision surface reconstruction based on a variational framework and the L1-norm; finally, utilizing an augmented Lagrangian method and a particle swarm optimization method to iterate and solve the vertex position and the sharpness of the control mesh of a subdivision surface, thereby obtaining the reconstructed subdivision surface preserving the features. Compared with the conventional method, the subdivision surface reconstruction method based on the variational feature perception provided by the present invention has the obvious advantages of preserving the features of the surface, such as the sharp features, the semi-sharp features, etc., can process the reconstruction problem of the input meshes with the noise of different scales, and has the wide application prospect on the surface reconstruction.