基于多孔多方向小波的SAR图像正则化超分辨重构

Regularied SAR image super-resolution reconstruction based on àtrous-directional wavelet

针对SAR图像超分辨重构问题,建立了基于多孔多方向小波域的正则化模型。在选取正则化参数时,提出一种自适应确定方法,该方法无需知道噪声大小和图像的先验知识,提高了确定正则化参数的准确性;求解模型时用FR共轭梯度法来改善算法的收敛性。最后将该算法分别与空域中正则化算法和小波域及轮廓波域中正则化算法进行了比较,仿真实验结果表明,该算法较好地再现了各种边缘信息,其重构结果均优于其他三种方法。

Aiming at the reconstruction of SAR image super-resolution, a regularization model based on atrous-nonsubsampied contourlet domain is established.When choosing the regularization parameter, an adaptive method is proposed which needn＇t the nosie and prior information of image, and enhances the veracity of the regularization parameter.The algorithm＇s convergence is improved by FR conjugate gradient method.Compared with regularization algorithm in spatial domain, wavelet and contourlet domain,computer simulations show that the proposed approach in this paper can properly retrieve the main information of original image and be superior to the other three methods.