Face Inpainting Combining Structured Forest Edge Information and Gated Convolution
Fuping Wang, Wenlou Li, Ying Liu, Yanchao Gong, Ziming Gao, Jin Lu · 2021
For the face inpainting under arbitrary shape occlusion, the existing methods are easy to produce edge blur and distortion of the inpainting results. In this paper, an algorithm for face inpainting combining structured forest edge information and gated convolution is proposed. Firstly, the edge contour of the occluded area is reconstructed by prior face knowledge to constrain the process of face inpainting. Secondary, the gated convolution holds the ability to extract accurate local feature when some pixels were missed, then a gated convolution based Generative Adversarial Network (GAN) for face inpainting is designed. The model consists of two parts: edge connection network and image inpainting network. The edge connection network accomplish the automatic completion and connection of the missing edge image. The image inpainting network takes the completed edge image as the guidance information, and combines the occlusion image to repair the missing face area. Compared with others, the experimental results show that the proposed algorithm has more precise detail information and better visual quality.