Cloud Shadows Detection and Compensation Algorithm on Multispectral Satellite Images for Agricultural Regions

Abstract

The presence of cloud shadows on remotely sensed images significantly complicates the analysis of the monitored area. The paper considers the problem of cloud shadows compensation on multispectral remotely sensed data. A new algorithm for cloud shadows detection and compensation based on a robust estimate of a local shadowing coefficient is proposed. Experimental results on shadow compensation quality for RGBN channels and Normalized Difference Vegetation Index (NDVI) index using the dataset of ten Sentinel-2 satellite multispectral images are presented. The results show that the compensation effect by the proposed algorithm on RGBN and NDVI data is 2 times better than that of the Gray-World-based algorithm.