Cloud Detection in High-Resolution Satellite Images

Detecting the cloud covarage of a satellite image is crucial for determining the image quality as well as preventing a perfromance loose of automatic detection algorithms because of the clouds. In this work, a new approach for detecting clouds in high-resolution satellite images has been proposed. The proposed method initially subsets the image into small patches and performs analyses based on edge and color information. Homogenous Texture Descriptors are then extracted for the patches that have the potential to contain cloud areas and the final classification of the patch is performed by using a Support Vector Machine (SVM) type classifier.