Despeckling Sentinel-1 GRD images by deep learning and application to narrow river segmentation
This work addresses the problem of speckle interference in satellite imagery for remote sensing applications, offering an incremental improvement in river segmentation.
The paper tackled speckle noise in Sentinel-1 GRD images using a self-supervised deep learning method based on SAR2SAR, resulting in improved detection of narrow rivers with better segmentation when applied as a pre-processing step.
This paper presents a despeckling method for Sentinel-1 GRD images based on the recently proposed framework "SAR2SAR": a self-supervised training strategy. Training the deep neural network on collections of Sentinel 1 GRD images leads to a despeckling algorithm that is robust to space-variant spatial correlations of speckle. Despeckled images improve the detection of structures like narrow rivers. We apply a detector based on exogenous information and a linear features detector and show that rivers are better segmented when the processing chain is applied to images pre-processed by our despeckling neural network.