Edge repair and bicubic interpolation-based enhancement for image restoration

Image restoration is a technique that reconstructs missing or damaged parts of an image by predicting the content in those areas. It is widely applicable in everyday scenarios and remains a prominent topic in computer vision. Despite its challenges, the advent of deep learning has significantly expanded the possibilities of image restoration. However, restored images often appear unnatural or show noticeable artifacts, particularly when dealing with large missing areas. To tackle these challenges, this paper introduces a novel image restoration model, E2I-muGIF, which integrates edge information into a convolution-based GAN restoration network to enhance post-restoration image quality. Extensive experiments on datasets such as CelebA, Places2, and ImageNet clearly demonstrate that our approach achieves superior visual performance.