Document Image Restore via SPADE-Based Super-Resolution Network

With the development of deep learning technology, various structures and research methods for super-resolution restoration natural images document have been introduced. In particular, a number recent studies conducted developed in image using generative adversarial networks. Super-resolution is an ill-posed problem because some complex restraints, such as many high-resolution being restored same low-resolution image, well difficulty restoring noises edges, light smudging, blurring. this study, we applied to text spatially adaptive denormalization (SPADE) structure, different from previous methods. This paper used SPADE solve problems edge unclearness, hardness catch features texts, color transition. As result it can be confirmed that character ambiguous stroke are more clearly when contrasting with other previously suggested Additionally, proposed method’s PSNR SSIM scores 8% 15% higher compared