An Algorithm to Remove Thin Clouds But to Preserve Ground Features in Visible Bands

Modeled thin-cloud components using the RTM-based algorithm may include the reflectance of ground targets. If the components are then used in cloud removal, the ground reflectance can be adversely affected. Thus, we revised the modeled components and developed a thin-cloud removal algorithm. With Landsat-8 data acquired on 29 March 2014, the spatial correlation coefficient between Landsat-8 Band-9 and revised cloud components was the highest. The revision of the modeled components was quantitatively satisfactory. The spatial correlation coefficients for Bands 1-4 between the cloud-removed and cloud-free reference images ranged from 0.687 to 0.797. In addition, the algorithm had the smallest impact on the cloud-free pixels. Overall, this algorithm outperformed the original RTM-based algorithm. Therefore, we have achieved the objective to remove thin clouds but to preserve ground features.