Oil spill detection in SAR imagery

The possibilities for automatic segmentation of oil slicks and oil slick look-alikes in ERS-1 SAR imagery is investigated. "Spots" in the images are detected using adaptive thresholding and statistical classification of oil slick candidates is performed using geometric properties and gradient-based features. Promising results are shown, but the methods need to be tested on a larger data set before any final conclusions can be drawn.<
>