Automatic Cloud Detection Using Spectral Rationing and Fuzzy Clustering

Cloud detection is important in many applications and it is an inevitable pre processing task before cloud removal and using satellite images for further applications such as numerical weather forecast, surface identification, change detection etc. Remote sensing images are partially unusable because of the presence of clouds especially images from humid tropical regions. The accurate detection of clouds in satellite images is vital for many applications. In this paper, we propose a novel algorithm for automatic cloud detection in Land sat images. This method can automatically detect clouds in satellite images without any manual interference. The proposed cloud detection algorithm performs color transformation on the input image, then generates a ratio image using spectral image rationing technique and finally clusters the ratio image using Fuzzy C-means clustering for detecting clouds automatically. The spectral rationing technique uses the ratio value of the croma and the luma to construct the ratio image for detecting clouds in satellite images. The pixels in cloud regions have lower values in the ratio image than those pixels in non cloud regions. This property can be used for detecting clouds. The experimental results demonstrate that proposed method is efficient in detecting thick clouds and thin clouds. Result analysis shows that the proposed algorithm correctly detect the clouds in average time.