基于l^k范数的正则化方法及其在SAR图像处理中的应用

Regularization method based on l^k norm and its application

本文基于lk范数与稀疏性之间的关系,提出了一种基于lk范数的正则化方法,该方法在抑制信号噪声的同时, 实现了信息的高精度稀疏表示。另外,通过深入研究基于lk范数的正则化方法,本文发现了正则化方法与广义岭估计的契合之处,提出了一种新的正则化参数的选取方法,解决了正则化参数选取的瓶颈问题。文中将该方法应用于SAR图像处理的实验,实验结果表明该方法具有有效的信息稀疏表示寻优功能。

Based on the relation between sparsity and l^k norm, a regularization method is presented for the sparse representation of signals. Simultaneously, the paper finds the agreement between the regularization and the ridge estimation, which affords a novel method of choosing the optimal regularization parameters. Experimental results of SAR images demonstrate that our method can find rapidly the optical regularization parameters and obtain accurately the sparse signal representation at the same time.