Spatial Resolution Enhancement for Halftone Images Using Convolutional Neural Networks

To print a grayscale image on a printer with current technology (laser or inkjet), it is necessary to convert it into a binary image called halftone, so that when a human being looks at it from a distance, the binary image visually resembles the original image. In this work, we present a technique to increase the spatial resolution of halftone images based on convolutional neural networks (CNNs). To our knowledge, this is the first work that seeks to solve this problem using CNNs. We use our algorithm to increase the resolution of binary halftones and show that our algorithm is considerably better the previous decision-tree based method. We also use our algorithm to increase the resolution of scanned old comics pages, and show that our algorithm is better than conventional resampling methods.