Information Theoretical Assessment of Methods for Segmentation of High Resolution Remote Sensing Images

Scene understanding of remotely sensed images requires a certain amount of preprocessing in order to remove, or alleviate the effects of, all those factors that disturb the imaging process. These factors depend essentially on the peculiar way in which each kind of sensor acquires the image (sensorrelated factors) and on the terrain topography, the illumination and the view angle (radiometric factors). In this paper, a Bayesian model-based maximum a posteriori estimation approach to correct these disturbing factors is suggested.