Structure preserving sar image despeck- ling via l0-minimization

Abstract

In this paper, we propose a new method for Synthetic Aperture Radar (SAR) image despeckling via L0-minimization strategy, which aims to smooth homogeneous areas while preserving significant structures in SAR images. We argue that the gradients of the despeckled images are sparse and can be pursued by L0-norm minimization. We then formularize the despeckling of SAR images as a global L0 optimization problem with ratio-of-average operations. Namely, the number of pixels with ratio-of-average that are unequal to one is controlled to approximate prominent structures in a sparsity-control manner. Finally, a numerical algorithm is also employed to solve the L0 optimization problem. In contrast with existing SAR image despeckling approaches, this strategy is applied without necessity to consider the local features or structures. The performance of our method is tested on high resolution X-band SAR images. The experimental results show the effectiveness of the proposed method in SAR image filtering. It outperforms many typical despeckling techniques in terms of the equivalent-number-of-looks and the edge- preserve-index. It also has some advantages compared with the existing state-of-the-art despeckling filters.