Modified ROA algorithm applied to extract linear features in SAR images

Linear feature is one of the important characteristics in SAR images because some objects manifest linear features, such as roads and bridges. However, linear feature detection in SAR images is not as easy as that in optic images because of the multiplicative speckle noise in SAR images. In order to extract linear features effectively, the typical edge detection algorithm in SAR images - ROA (ratio-of-averages) algorithm is modified in this paper, using the improved Hough transform. The result shows that this modified ROA algorithm is valid and efficient because a majority of linear features can be detected, while noise and fake edges can be restrained effectively