Single-Image Digital SuperResolution A Revised Gerchberg Papoulis Algorithm

Abstract—The subject of extracting high-resolution data from low-resolution images is one of the most important digital processing applications in recent years, attracting much research. In this work the authors show how to improve the resolution of an image when a small part of the image is given in high-resolution. To obtain this result the authors use an iterative procedure imposing the low frequencies complete data of the original low-resolution image and the high-resolution data present only in a fraction of the image. The procedure is based on the Gerchberg-Papoulis algorithm and contains dynamic properties, not present in the original scheme. The result is a clearer image, with higher correlation to the required high-resolution image. The authors show the use of such a procedure on a rosette image and on a facial image to demonstrate the higher frequencies obtained and on a text sample to show improvement in textual understanding. Index Terms—Image Processing, Signal reconstruction