A Morphological Approach to Astronomical Image Registration and Super Resolution Enhancement

Super resolution (SR) methods strive to generate a high resolution (HR) image utilizing several low resolution (LR) images taken from different views. Although this procedure is worthwhile in astronomical imagery, most of methods – including Harris feature point extraction – are not appropriate for registering astronomical images. In this work, we propose a simple, fast and accurate registration method in order to derive suitable feature points based on morphological image processing applicable to astronomical images due to their characteristics. Subsequently, we use the set of obtained feature points as an input to RANSAC for deriving the transformation matrices for each image. Prior to super resolution, registered LR images are projected onto a HR grid. After determining a specific Point Spread Function (PSF) for atmospheric blur, by utilizing an Adaptive Wiener Filter iteratively updated using Steepest Descent approach, filter weights are specified to interpolate intensities throughout the grid. Based on quantitative simulation results provided, the proposed registration and super-resolution methods result in higher reconstruction quality in a PSNR sense, in comparison with common methods.