arxivst stuff from arxiv that you should probably bookmark

SAR image despeckling through convolutional neural networks

Abstract · Apr 2, 2017 09:44 ·

cs-cv

Arxiv Abstract

  • G. Chierchia
  • D. Cozzolino
  • G. Poggi
  • L. Verdoliva

In this paper we investigate the use of discriminative model learning through Convolutional Neural Networks (CNNs) for SAR image despeckling. The network uses a residual learning strategy, hence it does not recover the filtered image, but the speckle component, which is then subtracted from the noisy one. Training is carried out by considering a large multitemporal SAR image properly despeckled through 3D filtering, in order to approximate a {\em clean} image. Experimental results, both on synthetic and real SAR data, show the method to achieve performance that improve with respect to state-of-the-art techniques.

Read the paper (pdf) »