Thick clouds, thin clouds, and cloud shadows inevitably degrade high spatial resolution optical remote sensing images (HRIs), thereby greatly restricting their subsequent application. Thin and thick clouds often coexist in HRIs, and the existing methods rarely achieve integrated removal. In this paper, we propose a joint thin–thick cloud removal (JTTR) framework to effectively remove both types of cloud contamination. The JTTR framework employs a multi-scale cloud detection approach to automatically select clean samples for the estimation of a thin cloud thickness map. In addition, an information cloning model based on radiometric adjustment and boundary interpolation (ICM-RABI) is incorporated. For the first time, the ICM-RABI model was applied to thin cloud removal, enabling joint removal of both thin and thick clouds. The proposed JTTR framework achieved a satisfactory performance across multiple sensors and complex scenes in both real and simulated experiments using Gaofen satellite data, demonstrating its strong generalization ability and great potential for practical applications.
Zhu et al. (Wed,) studied this question.