Figure Patches Learning-based Super-Resolution

Abstract

In this paper, we describe the patch learning-based super-resolution using simple pattern figure images (ex. circle) rather than natural images to learn image patches. In the learning-based super-resolution, if we use natural images for learning, we have to consider whether the image characteristics are suitable and how many images are needed to learn. However, by using simple pattern figure image learning we overcome these difficulties and enables good restoration in result images.