Nonlocal Model-Free Denoising Algorithm for Single- and Multichannel SAR Data

Abstract

Among the large number of Synthetic Aperture Radar (SAR) image despeckling approaches existing in literature, Non-Local (NL) filters have received a desirable boost. However, often NL approaches define the similarity critirion based on model assumptions, such as a fully developed speckle model. This assumption may not be verified in high-resolution images of urban environments. To address this issue, a stand-alone model-free despeckling framework is proposed in this paper. The presented approach provides a generic framework for denoising a variety of SAR products, from a single intensity/amplitude image to polarimetric and interferometric SAR data. In particular, the method is based on the empirical distributional similarity between the patch containing the pixel to be recovered and the patch containing a similar candidate pixel. To decide whether the patches follow a similar distribution, the Kolmogorov-Smirnov test is adapted. Finally, the restoration process aggregates the selected similar pixels based on their relative importance derived from their distribution similarities. To mitigate the blurring effect and preserve the resolution, the inhomogeneity of the ratio image is used to perform the bias reduction step. The designed generic despeckling filter was tested on different products of SAR data. The results show that the method proves to be an unbiased restoration approach and is able to preserve structures and textures. It works fully automatically and efficiently with single and multi-look (and multi-channel) images.

Fast Patchwise Nonlocal SAR Image Despeckling Using Joint Intensity and Structure Measures

“…In this section, GNLMS is derived from GNLM as only SAR images considered for a fair comparison. (4) MEET-SAR [34]: A model-free nonlocal denoising method, which is independent of the noise model. For all of the reference methods, the suggested parameters in the original papers were retained.…”

“…Most of these methods rely on strict speckle noise model assumption, while the assumption cannot be verified in some cases. Hence, a model-free nonlocal approach was proposed with two similarity criteria, one of which is obtained by directly comparing the distributions of the patches, while the other compares the distribution of ratio patches [34].…”

Speckle Reduction in SAR Images

“…The covariance matrix was defined from the patch information in [53], while the generalized likelihood ratio test was used following the distribution of the covariance matrix as the similarity index. In [11,14], the statistical distribution strategy was extended and the patch ratio was compared with a known model. The details of this model are described in the next subsection.…”

“…Unlike a local method such as the well-known intensity-driven adaptive neighborhood (IDAN) filter [10] that considers connected pixels, far-spaced pixels whose patch information is similar to the patch of that target pixel can be combined, justifying the widely used nonlocal name. In the last decade, several NL algorithms [11][12][13][14][15][16] for despeckling have appeared in the literature. The main difference between the algorithms lies in the definition of the similarity criterion and the function used to merge similar pixels or patches.…”

Synthetic Aperture Radar Image Despeckling Based on a Deep Learning Network Employing Frequency Domain Decomposition

“…This model also used an updated despeckling gain loss function. In addition, Aghababaei et al [35] proposed a stand-alone model-free non-local (NL) despeckling framework for the removal of speckle noise from single-channel and multi-channel SAR data. In addition to the above-mentioned fully supervised training methods, many semi-supervised and self-supervised training methods have been proposed, such as SAR2SAR [36], Speckle2Void [37], and MERLIN [38].…”