Super Resolution Using Neural Network

Abstract

Super resolution imaging refers to inferring the missing high resolution image from low resolution image(s). Super resolution methods are generally classified into reconstruction based and learning based methods. The learning based methods fully learn the intensity prior between all bands of images. For most of these, some fake information can be inevitably introduced into synthesized high resolution image though the image may offer visually good effects. For medical image analysis, it might deteriorate the image analysis and diagnosis performance. In this paper we propose a better learning using real image training set that enhances the high frequency information. The method exploits richness of real-world images. The training set is preprocessed images so as to extract the structural correlation. The technique learns the fine details that correspond to different image structures seen at a low-resolution and then uses those learned relationships to predict fine details in other images.