Abstract:
This paper presents a new algorithm for SAR images segmentation based on
thresholding technique. Histogram of SAR images is assumed to be a combination of
Gamma distribution. The maximum-likehood technique is therefore used to estimate
the histogram parameters. Thresholds are selected at the valleys of a multi-modal
histogram by minimizing the discrimination error between the classes of pixels in
the image. When this algorithm is tested on articial histograms, an accurate
estimate of their parameters is obtained. The algorithm is applied to several
RADARSAT SAR images, with the number of looks for each image being 4 or 8. The
results obtained are promising.