Local Regression Model for Automatic Face Sketch Generation

This paper proposes an automatic face sketch generation approach which learns the mapping relationship from photo-sketch pair examples through the local regression model.The proposed method takes into account both face fidelity and aesthetics.First, the relationship between a face photo and its corresponding sketch is learned on image patch level by exploring some regression techniques.Then, by applying the learned relationship to the input face photo, the final face sketch is generated with the consideration of structural information and the simulation of pencil stroke texture.Experiments conducted on CUFS database have shown that our results are more appealing than previous methods for preserving of both face identity and aesthetics.