局部回归模型驱动的人脸素描自动生成

Local Regression Model for Automatic Face Sketch Generation

为了兼顾生成人脸素描与对应照片间的相像性和艺术性,提出一种利用局部回归模型学习照片-素描间的映射关系自动生成人脸素描的方法.首先在图像局部区域上采用回归技术获得照片-素描对应局部区域之间的映射关系;然后将映射关系应用到输入的人脸照片上,考虑结构性特征并模拟铅笔素描笔触来合成对应的素描图像.在CUFS数据库上的实验结果表明,采用该方法合成的素描保持了原始人脸照片的身份信息,并且体现出较好的铅笔素描纹理.

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.