Damaged Old Photo Inpainting Algorithm Based on Reverse Mask and WGAN-GP
Aiming at the problems of blur and artifacts in old photo inpainted by current old photo repairing algorithms,a dam-aged old photo inpainting algorithm that combines the reverse mask mechanism and generative adversarial network is proposed.This method utilizes reverse mask techniques focused on the target area to be inpainted,and proposes a joint perception-driven loss func-tion and WGAN-GP discriminator network training to enhance the global and local consistency of the inpainted image.In addition,by transforming the RGB image color space and adding Gaussian filtering,the old photo image data set is expanded.After experi-ments,the results show that the algorithm in this paper can accurately inpaint the missing parts in old photo compared to the compar-ison algorithm.The structure and texture of the inpainted photo are consistent with the original photo,which is more natural in the senses and vision.
old photoimage inpaintinggenerative adversarial networkreverse maskgradient penalty
国家科技期刊平台
NSTL
万方数据
