Automatic Delineation of Clouds and Their Shadows in Landsat and CBERS (HRCC) Data

The presence of clouds and their shadows is an obvious problem for maps obtained from multispectral images. As a matter of fact, clouds and their shadows create occluded and obscured areas, hence information gaps that need to be filled. The usual approach-pixel substitution-requires first to recognize the cloud/shadow pixels. This work presents a cloud/shadow delineation algorithm, the cloud/shadow delineation tool (CSDT) designed for Landsat and CBERS medium resolution multispectral data. The algorithm uses a set of literature indices, as well as a set of mathematical operations on the spectral bands, in order to enhance the visibility of the cloud/shadow objects. The performance of CSDT was tested on a set of scenes from the Landsat and CBERS catalogues. The obtained results showed more accurate and stable performance on Landsat data. In order to validate the proposed approach, this work presents also a comparison with the F-mask algorithm on Landsat scenes. Results show that the F-mask technique tends to overestimate the cloud cover, while CSDT slightly underestimates it. However, accuracy measures show a significantly better performance of the proposed method than the F-mask algorithm in our investigation.