A fast multi-temporal optical data cloud removal method based on isophote constraint

Abstract

The problem of cloud cover significantly affects the quality and utility of optical satellite imagery, posing challenges for image analysis and interpretation. In response, this research presents a fast and efficient image reconstruction method by exploiting the strength of isophote in linear structure propagation of ground features, providing users with a practical option for cloud removal task. In addition, this research solves the inapplicability problem when the cloud-contaminated regions overlap between different temporal images, and then extends the application to the field of multi-temporal data cloud removal. Through several experiments, the proposed method outperforms other comparative methods for both dual-temporal data reconstruction and multi-temporal data reconstruction task. Furthermore, the method proves to be efficient and robust even with limited auxiliary information. The proposed multi-temporal image reconstruction framework is available at: https://github.com/YuXiaoyu221/ICMIR.