Automatic detection of clouds in multispectral images subjected to interchannel parallax

摘要

The present paper considers a multistage algorithm of the automatic detection and segmentation of clouds on images of the Earth surface acquired by scanners performing multi-temporary imagery in different spectral bands. High reliability of the cloud detection is reached by taking into account a stereoscopic effect becoming apparent in the mutual shift of cloud bounds. At the first stage all white color objects are selected. At the second stage selected objects are filtered by the parallax criterion. Such multistage approach allows increasing the algorithm performance. The paper contains evaluations of the cloud segmentation reliability and algorithm performance.