Minimum description length Shape Model based on bio-inspired features

This paper proposes an enhanced MDL Shape Model to solve the point correspondence problem. The current MDL methods build models mainly based on shape information and may get bad models. Motivated by the biologically inspired features (BIF), which inspired by visual cortex, we compute the C1 response on the master node and add the cost of BIF across training set to the objective function of MDL Shape Models. Experiments show that our method can get better model and point correspondence.