Removing cloud shadows from ground-based solar imagery

TL;DR

This paper introduces a novel U-Net based method, including a conditional GAN approach, to effectively remove cloud shadows from ground-based solar images, enhancing the visibility of solar structures for space weather analysis.

Contribution

It presents a new deep learning framework for cloud shadow removal in solar imagery, outperforming traditional and baseline methods across multiple datasets and metrics.

Findings

Improved image quality metrics over baseline methods

Effective removal of various cloud types and textures

Demonstrated on real and synthetic datasets

Abstract

The study and prediction of space weather entails the analysis of solar images showing structures of the Sun's atmosphere. When imaged from the Earth's ground, images may be polluted by terrestrial clouds which hinder the detection of solar structures. We propose a new method to remove cloud shadows, based on a U-Net architecture, and compare classical supervision with conditional GAN. We evaluate our method on two different imaging modalities, using both real images and a new dataset of synthetic clouds. Quantitative assessments are obtained through image quality indices (RMSE, PSNR, SSIM, and FID). We demonstrate improved results with regards to the traditional cloud removal technique and a sparse coding baseline, on different cloud types and textures.