Old Document Restoration using Super Resolution GAN and Semantic Image Inpainting
Yong Jun Kim, Debapriya Hazra, Yung-Cheol Byun, Khi-Jung Ahn · 2019
Restoration of damaged images is a fundamental problem that has been attempted before the advent of digital image processing technology. In this paper, one of the deep neural network technologies (GAN), we propose an image restoration network using Generative Adversarial Network. The proposed system is the image generation network, the generation result plateIt consists of a star network. Old documents not only contain information, but also we can learn about historical people's thought and consciousness from the past. Old document restoration is referred to as the restoration of documents that are usually made of parchment which are damaged either naturally or artificially. Missing regions in old documents are filled based on the current visual data which is a hard task in image inpainting. In this paper, we present Super Resolution Generative Adversarial Network (SRGAN) and semantic image inpainting for restoring the old documents so that they can be reused.