L1/2 Regularization Sar Imaging Via Complex Image Data: Regularization Parameter Selection for Target Detection Task

In this paper, we proposed an L
_1/2
regularization SAR imaging method via complex SAR images. Compared with L
₁
regularization method, our approach can better reconstruct the target with higher grayscale value, which is very important for target detection. In addition, we proposed a novel regularization parameter selection method for target detection task. First, we established the relationship between the regularization parameter and the scene sparsity. Then, by using a square window detector to estimate the scene sparsity, the regularization parameter is iteratively updated to suppress the background clutter at the greatest extent while retain the majority of the target area. Experimental results demonstrated that the images reconstructed by the proposed method can gain better target detection results than traditional imaging method.