Unsupervised Change Detection in Satellite Images using Oversegmentation and Mutual Information

Taskesen, Bahar
Besbinar, Beril
Koz, Alper
Alatan, Abdullah Aydın
In this paper, a novel solution to the problem of unsupervised change detection in bitemporal satellite images is presented. Information measures, which are well-known and commonly-used in the change detection literature, result in unsharp change maps and masks without well defined boundaries as a result of local computation. In the proposed method, mutual information with local joint distributions computed within the over-segments after image registration, radiometric correction and some preprocessing steps are observed to eliminate the problem of sharpness. Results, which are presented comparatively with fundamental approaches, show that the change masks obtained by the proposed method are convenient for different application areas, such as damage assesment of man made structures after natural disasters, and/or urban planning.
Citation Formats
B. Taskesen, B. Besbinar, A. Koz, and A. A. Alatan, “Unsupervised Change Detection in Satellite Images using Oversegmentation and Mutual Information,” presented at the 25th Signal Processing and Communications Applications Conference (SIU), Antalya, TURKEY, 2017, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/55940.