Unsupervised change detection using a novel fuzzy c-means clustering simultaneously incorporating local and global information

This paper presents a novel fuzzy c-means (FCM) clustering simultaneously incorporating local and global information (FLGICM) method to unsupervised change detection (CD) from remotely sensed images. A new factor including three local, global and edge parameters is added into the conventional FCM to enhance the insensitivity to noise and preserve detailed features. The spatial attraction between the central pixel and its neighborhood pixels is incorporated as a local parameter to utilize spatial information. A global parameter designed based on the estimated mean values of changed and unchanged pixels is introduced into the new factor to enhance its robustness and ability of separating changed from unchanged pixels. In addition, an edge parameter is also added to remain accurate edges and change details. Two experiments were carried out on Landsat images to test the performance of FLGICM. Experimental results indicate that FLGICM always achieves high accuracy and overperforms some state-of-the-art CD methods. Therefore, the proposed FLGIC provides an effective unsupervised CD method.