Segment-based Remote Sensing Image Change Detection Using Normalized Mutual Information

This paper proposed a change detection method,which uses the segments obtained from the image segmentation.Normalized mutual information was introduced into the change detection of remote sensing images.Histogram and joint histogram of segments are used to calculate the normalized mutual information.The best threshold was calculated based on the smallest conditional entropy.Using the normalized mutual information the method compared with the one using correlation coefficient,it shows the normalized mutual information is more suitable for change detection of remote sensing images.