Residual Enhancement Network for Realistic Face Sketch-Photo Synthesis

Abstract

Face sketch-photo synthesis is a significant challenge task in computer vision area, due to the blurred facial details and color distortion produced by the existing approaches. In this paper, we propose a realistic face sketch-photo synthesis method based on residual enhancement network. In the network, a residual enhancement module is constructed and embedded in U-Net to improve the feature representation capability of the deep network. In addition, a detail loss and a perception loss are adopted to constrain the synthesized image has abundant detail and realistic photo style. Experimental results on multiple face sketch datasets indicate that the proposed method obtains superior performance than the state-of-the-art methods, both in terms of visual perception and objective evaluations.