Cloud detection from satellite images based on Haar wavelet and clustering

Cloud covers normally present in satellite images which is one of the major obstacle in the processing of satellite images. Remote sensing images from tropical regions are partially unusable especially because of the presence of clouds. For improving the accuracy of land cover classification in satellite images cloud detection is useful. The major obstacle in cloud detection approach is the misclassification rate of cloud pixels. In this paper novel cloud detection from satellite image based on stationary Haar wavelet is presented. The proposed approach uses wavelet transform, histogram analysis and clustering for cloud detection. Remote sensing satellite images can be considered as a signal. Clouds are low frequency components in satellite image. So wavelet analysis of satellite images is helpful to extract low frequency components and detection of clouds. The result analysis shows that the proposed method can detect clouds automatically from satellite images with high accuracy. Experimental analysis is performed on Landsat-7 Enhanced Thematic Mapper Plus (ETM+) sensor images and results are obtained.