Column fixed-pattern noise removal in solar images using two-way filtering

Abstract

Solar images are critically important for studying solar activities and features. Today, many observatories rely on CMOS sensors to acquire these images. However, these sensors often introduce column fixed-pattern noise (CFPN), seriously affecting image quality. Therefore, we proposed a two-way filtering algorithm to remove CFPN. Firstly, in the horizontal direction, we used the one-dimensional global weighted least squares filter and the efficient bilateral filter to obtain a coarse denoised image. Then, we utilized the weighted guided filter in the vertical direction to estimate the CFPN components, thereby obtaining a clean solar image. We selected three different solar observation images to compare and evaluate our results to those obtained by three comparative methods. The images are observed by the Solar Upper Transition Region Imager aboard the SATech-01 satellite. Additionally, we further used two quantitative metrics, photo response non-uniformity and mean relative deviation, to quantify the denoised results. The results demonstrate that our proposed method removes the CFPN better and preserves the image features in a more balanced way.