SAR Image Segmentation Exploiting No Background Knowledge on Speckled Radiance: A Feasibility Study

This work presents a SAR image segmentation scheme consisting of a sequence of four modules, all selected from the literature. These modules are: i) a speckle model-free contour detector that is the core of the segmentation scheme; ii) a geometrical procedure to detect closed regions from non-connected contours; iii) a region growing procedure whose merging rules exploit local image properties, both topological and spectral, to eliminate artifacts and reduce oversegmentation introduced by the second stage; iv) a neural network clustering algorithm to detect global image regularities in the sequence of within-segment properties extracted from the partitioned image provided by the third stage. In the framework of a commercial image-processing software toolbox, the proposed SAR image segmentation scheme employs a contour detector that is promising because: i) it is easy to use, requiring the user to select only one contrast threshold as a relative number; and ii) it exploits no prior domain-specific knowledge about the data source and the content of the scene, i.e., it is capable of processing SAR images as well as both achromatic and multi-spectral optical images. The segmentation scheme is tested on three images acquired by different SAR sensors. The robustness of the segmentation method is assessed by changing only one parameter of the procedure in the different experiments. Experimental results are interpreted as an encouragement to focus further multidisciplinary research on how to combine responses of multi-scale filter banks in low-level visual systems.

ICSI Technical Report TR-98-018

Technical Report