The Automatic Segmentation of High Resolution Synthetic Aperture Radar Stripmap Imagery of Natural and Mixed Urban Terrains

J.M. Henson, M.C. Grund, and W.P. Waite (USA)


Radar, Image, Segmentation.


The following paper describes techniques for the segmentation of high resolution SAR stripmap imagery of naturally occurring terrains with urban features. A discussion of the complex raw data and image formation techniques is also presented. Prior to image segmentation a number of fading noise reduction filtering techniques are applied. The basic segmentation algorithm consists of the use of image histogram derivatives to select gray level thresholds. These thresholds are used consecutively to produce a series of binary images. Each binary image is segmented by the independent application of a dilation and an erosion algorithm. The results are subsequently subtracted to form a series of segmented images, which can be added as desired to form a fully segmented composite image.

Important Links:

Go Back