Cloud removal using multi-temporal satellite images and fuzzy logic [ READ ]
Soumya Sucharita, Satyajit Mohanty, Manoj Pandya, M. B. potdar, Leena Patel, Krunal patel
Satellite images are often obscured by cloud cover and shadows. This poses a major challenge to processing of data regarding the surface underneath. Various methods have been proposed by different researchers to obtain cloud and cloud shadow free images from satellite images. In this paper, an approach is suggested which generates cloud free image from multi-temporal images. This paper briefly reviews the existing methods for reconstruction of cloud contaminated images and then suggests an approach based on segmentation of image using Fuzzy logic. The cloud contaminated patches are detected in an image using Fuzzy C means algorithm. Following this step, blobs are detected and those pixels are replaced with the corresponding pixels from the cloud free image under the assumption that the land cover changes insignificantly during a short period of time. Filtering technique is used to remove the visible seams in the reconstructed image. Experimental analysis is conducted on satellite images and results are obtained. Both thin and thick clouds are removed effectively. We have also compared the performance of various segmentation algorithms in cloud detection. This proves the improved performance with the proposed approach.
Soumya Sucharita, Satyajit Mohanty, Manoj Pandya, M. B. potdar, Leena Patel, Krunal patel
Satellite images are often obscured by cloud cover and shadows. This poses a major challenge to processing of data regarding the surface underneath. Various methods have been proposed by different researchers to obtain cloud and cloud shadow free images from satellite images. In this paper, an approach is suggested which generates cloud free image from multi-temporal images. This paper briefly reviews the existing methods for reconstruction of cloud contaminated images and then suggests an approach based on segmentation of image using Fuzzy logic. The cloud contaminated patches are detected in an image using Fuzzy C means algorithm. Following this step, blobs are detected and those pixels are replaced with the corresponding pixels from the cloud free image under the assumption that the land cover changes insignificantly during a short period of time. Filtering technique is used to remove the visible seams in the reconstructed image. Experimental analysis is conducted on satellite images and results are obtained. Both thin and thick clouds are removed effectively. We have also compared the performance of various segmentation algorithms in cloud detection. This proves the improved performance with the proposed approach.
