Local motion contour segmentation algorithm of SAR image based on ROEWA operator

Aiming at the problem that the classical region-scalable fitting (RSF) model cannot segment the complex synthetic aperture radar (SAR) image, an improved RSF local active contour model based on the ratio of exponentially weighted averages (ROEWA) operator is proposed. The algorithm first uses the negative exponential function of the ROEWA operator of the SAR target edge as an edge indicator function to weight the region-scalable energy term, eliminating the edge detection failure of the gradient operator for the SAR image with multiplicative coherent speckles in the original model. Meanwhile, the method also prevents the leakage of the target weak boundary and avoids the lack of the whole boundary. Then the area term with variable weight coefficient is added to the partial differential equation with zero level set evolution to improve the adaptive capture ability of the algorithm around the target edge. At the same time, it also improves the ability of the algorithm to detect the target multi-layer contour, and preserves the outer contour of the target to the utmost extent. The variable weight coefficient of the area item can be automatically resized according to the modulus of the target's ROEWA operator, which keeps the edge details of the target very well. The improved algorithm is not affected by the initial contourand, insensitive to speckle noise. Experiments show that the computational complexity of the algorithm is only related to the size of the image. Through experiments on synthetic and real SAR image data, the intuition and validity of the proposed method are proved. © 2019, Editorial Office of Systems Engineering and Electronics. All right reserved.