Enhancing the Resolution of Satellite Images Using the Best Matching Image Fragment

摘要

Due to very high costs and a long revisit time, it is challenging to obtain good quality satellite images of the area of interest. As a result, super resolution reconstruction (SRR) methods which allow for creating a high-resolution (HR) image based on single or multiple low-resolution (LR) observations are being extensively developed. In this paper, we propose a few improvements to well-known single-image SRR technique based on a dictionary of pairs of matched LR and HR image fragments. The modifications concern both increasing the number of pairs of images fragments and the reconstruction algorithm itself in order to achieve visually pleasing results. This allows us to increase the quality of newly produced HR satellite images what is supported by conducted experiments.