Marker-controlled SAR image segmentation algorithm

This paper presents an algorithm for automatic segmentation of SAR imagery into target, shadow, and background clutter regions, which is a key pre-processing step for SAR image based automatic target recognition. Power transformation is employed in this algorithm to transform SAR image data to nearly Gaussian distributed data, followed by which is a de-nosing processing by using partial differentiation equation. A marker-controlled based SAR image segmentation algorithm is used to segment the de-noised images. We introduce a new sliding window method to find extreme value in SAR image, which is used as marker for controlling segmentation region. Experiment results show that this algorithm can overcome the influence of impulse noise and the correlation and dependency of adjacent pixel in SAR imagery, and get trim borderline and good segmentation performance.