Validation of segmentation techniques for SAR images

Segmentation is a vital technique in describing the structure of SAR images and in combating the effects of speckle that dominate such images. The performances of five segmentation algorithms are assessed and compared using real and simulated SAR data. The 'best' algorithm is sought using objective measures based on segment homogeneity and shape measures. These measures are compared to theoretical values and visual impressions of the images.