Despecking of SAR images using compressive imaging framework
Mahboob Iqbal, Jie Chen · 2012
A novel technique for despeckling of synthetic aperture radar (SAR) is proposed. A predefined number of overlapping subsets of pixels are selected from SAR image. Each subset is comprised of pixels selected from uniformly distributed locations. The subsets of pixels are elected in such a way that at least 20% of pixels in any subset should be different from pixels in any other subset. By considering each subset as compressive samples, a complete SAR image is reconstructed using convex optimization algorithm. These compressive reconstructed images are used to obtain despeckled SAR image. The proposed technique is tested on patches from stripmap TerraSAR-x data set. The proposed despeckling outperforms other benchmark despeckling methods in terms of visual quality as well as despeckling capability measuring metrics.