On Automatic Threshold Selection in Regression Method for Change Detection in Remote Sensing Images

Automatic threshold selection in regression method of change detection in remote sensing images is investigated in this paper. We have reviewed the three well-known automatic threshold selection methods in the image processing field; namely, convex hull, Otsu, and ISODATA. We have then described the adaptive threshold selection concept. We have then shown that from these three methods, only the ISODATA can be successfully applied to regression method and have proposed an ISODATA-based adaptive threshold selection method. Experimental results show the efficiency of the proposed adaptive threshold selection method.