The use of data fusion for the classification of dense urban environments: The Mytilene case

An example of the great advantages provided by data fusion is described in this paper. This exploits the merging of high spatial resolution images with high spectral resolution ones to increase the sensitivity of the classification to the more subtle variations to be found in such environments.. A Landsat-TM image and a panchromatic (PAN) image of the town area of Mytilene were used. From a number of methods investigated (intensity-hue-saturation — IHS, principal components analysis — PCA, high-pass filter technique — HPF) the HPF image was selected for a maximum likelihood classification. The gain produced by data fusion was the recreation of high spatial frequency classes like roads, buildings etc. that were smoothed in the TM image but existed in the PAN image, and this enabled the classification of a dense urban environment, such as the small town of Mytilene, to be achieved for the first time.