基于l_k范数正则化方法的SAR图像超分辨

SAR Image Super-resolution Based on Regularization of l_k Norm

提高合成孔径雷达(SAR)图像的分辨率对自动目标识别等具有重要意义。为此改进了一种基于lk范数正则化方法,并用于SAR图像超分辨。该方法通过合理开发利用符合SAR成像工程背景的先验知识,构造附加约束,把图像超分辨问题规划为形式简单的带约束优化问题。仿真和实测数据计算结果证实了该方法的有效性。

Enhancing the resolution of SAR image can improve ATR significantly. At present, super-resolution methods of SAR image are mostly based on single frame of SAR image. SAR super-resolution mentioned in this paper indicates the postprocessing of SAR images that has been carried out the imaging formation. So that it differ very much from super-resolutlon methods of SAR imaging, such as modern spectral estimation techniques and data extrapolation techniques etc. One regularization method based on lk norm used for super-resolution processing of SAR image is discussed in this paper. The observation relationship model is established in the complex image domain. So there is no need to construct the SAR projection operator. Based on parameter selection of Tikhonov regularization, Automatic optimal regnlarization parameter selection method of lk norm Regularization is also proposed. Regularization method based on lk norm exploits the useful prior information which is well consistent to the actual background of SAR imaging, makes up the additional constraint condition which implied sparseness constraint, turns the problem of super-resolution processing of SAR image into the simple-formed constrained optimization problem. The Simulation results and real data computation proves its validity.