# As if by magic: self-supervised training of deep despeckling networks with MERLIN

arXiv: https://arxiv.org/abs/2110.13148 · HTML: https://mlsanity.com/paper/2110.13148/

## Key facts

- Cited by 18 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 55 references that are in the map.
- First posted on arXiv 25 Oct 2021.

## Builds on (most cited first, 16 of 16)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Noise2Noise: Learning Image Restoration without Clean Data](https://mlsanity.com/paper/1803.04189/) (2018)
- [Why gradient clipping accelerates training: A theoretical justification for adaptivity](https://mlsanity.com/paper/1905.11881/) (2019)
- [High-Quality Self-Supervised Deep Image Denoising](https://mlsanity.com/paper/1901.10277/) (2019)
- [Deep Learning Meets SAR](https://mlsanity.com/paper/2006.10027/) (2020)
- [Noise2Same: Optimizing A Self-Supervised Bound for Image Denoising](https://mlsanity.com/paper/2010.11971/) (2020)
- [SAR image despeckling through convolutional neural networks](https://mlsanity.com/paper/1704.00275/) (2017)
- [Speckle2Void: Deep Self-Supervised SAR Despeckling with Blind-Spot Convolutional Neural Networks](https://mlsanity.com/paper/2007.02075/) (2020)
- [SAR Image Despeckling Using a Convolutional Neural Network](https://mlsanity.com/paper/1706.00552/) (2017)
- [Deep Learning Methods For Synthetic Aperture Radar Image Despeckling: An Overview Of Trends And Perspectives](https://mlsanity.com/paper/2012.05508/) (2020)
- [SAR Image Despeckling by Deep Neural Networks: from a pre-trained model to an end-to-end training strategy](https://mlsanity.com/paper/2006.15559/) (2020)
- [Learning a Dilated Residual Network for SAR Image Despeckling](https://mlsanity.com/paper/1709.02898/) (2017)
- [Noise2Kernel: Adaptive Self-Supervised Blind Denoising using a Dilated Convolutional Kernel Architecture](https://mlsanity.com/paper/2012.03623/) (2020)
- [A review of deep-learning techniques for SAR image restoration](https://mlsanity.com/paper/2101.11852/) (2021)
- [Towards Deep Unsupervised SAR Despeckling with Blind-Spot Convolutional Neural Networks](https://mlsanity.com/paper/2001.05264/) (2020)
- [Speckle Reduction in Polarimetric SAR Imagery with Stochastic Distances and Nonlocal Means](https://mlsanity.com/paper/1304.4634/) (2013)

## Built on by (most cited first, 18 of 18)

- [OpenForest: A data catalogue for machine learning in forest monitoring](https://mlsanity.com/paper/2311.00277/) (2023)
- [SAR Despeckling via Regional Denoising Diffusion Probabilistic Model](https://mlsanity.com/paper/2401.03122/) (2024)
- [Self-Supervised Learning from Noisy and Incomplete Data](https://mlsanity.com/paper/2601.03244/) (2026)
- [Multi-temporal speckle reduction with self-supervised deep neural networks](https://mlsanity.com/paper/2207.11095/) (2022)
- [Deep Learning-Based Anomaly Detection in Synthetic Aperture Radar Imaging](https://mlsanity.com/paper/2210.16038/) (2022)
- [PolMERLIN: Self-Supervised Polarimetric Complex SAR Image Despeckling with Masked Networks](https://mlsanity.com/paper/2401.07503/) (2024)
- [Just Project! Multi-Channel Despeckling, the Easy Way](https://mlsanity.com/paper/2408.11531/) (2024)
- [Equivariant Denoisers for Image Restoration](https://mlsanity.com/paper/2412.05343/) (2024)
- [Fast strategies for multi-temporal speckle reduction of Sentinel-1 GRD images](https://mlsanity.com/paper/2207.11111/) (2022)
- [Shot Noise Reduction in Radiographic and Tomographic Multi-Channel Imaging with Self-Supervised Deep Learning](https://mlsanity.com/paper/2303.14429/) (2023)
- [Linear Attention Based Deep Nonlocal Means Filtering for Multiplicative Noise Removal](https://mlsanity.com/paper/2407.05087/) (2024)
- [Multi-task SAR Image Processing via GAN-based Unsupervised Manipulation](https://mlsanity.com/paper/2408.01553/) (2024)
- [Sim2Real SAR Image Restoration: Metadata-Driven Models for Joint Despeckling and Sidelobes Reduction](https://mlsanity.com/paper/2601.01541/) (2026)
- [A Deep Learning Iterative Framework for Sentinel-1 Stripmap Enhancement Based on Azimuth Doppler Decomposition](https://mlsanity.com/paper/2605.29088/) (2026)
- [γ-Bridge: A Look-Parametric Diffusion Bridge](https://mlsanity.com/paper/2607.22719/) (2026)
- [Implicit Neural Speckle Denoising](https://mlsanity.com/paper/2608.06574/) (2026)
- [Hardware-Aware Deployment of Joint SAR Compression and Despeckling on FPGA](https://mlsanity.com/paper/2608.11271/) (2026)
- [ProSR: Semantic-Prototype-Guided Discrete Modeling for Physically Consistent SAR Super-Resolution](https://mlsanity.com/paper/2609.02377/) (2026)

## Abstract (arXiv)

Speckle fluctuations seriously limit the interpretability of synthetic aperture radar (SAR) images. Speckle reduction has thus been the subject of numerous works spanning at least four decades. Techniques based on deep neural networks have recently achieved a new level of performance in terms of SAR image restoration quality. Beyond the design of suitable network architectures or the selection of adequate loss functions, the construction of training sets is of uttermost importance. So far, most approaches have considered a supervised training strategy: the networks are trained to produce outputs as close as possible to speckle-free reference images. Speckle-free images are generally not available, which requires resorting to natural or optical images or the selection of stable areas in long time series to circumvent the lack of ground truth. Self-supervision, on the other hand, avoids the use of speckle-free images. We introduce a self-supervised strategy based on the separation of the real and imaginary parts of single-look complex SAR images, called MERLIN (coMplex sElf-supeRvised despeckLINg), and show that it offers a straightforward way to train all kinds of deep despeckling networks. Networks trained with MERLIN take into account the spatial correlations due to the SAR transfer function specific to a given sensor and imaging mode. By requiring only a single image, and possibly exploiting large archives, MERLIN opens the door to hassle-free as well as large-scale training of despeckling networks. The code of the trained models is made freely available at https://gitlab.telecom-paris.fr/RING/MERLIN.