English

Transformer-based SAR Image Despeckling

Computer Vision and Pattern Recognition 2022-01-25 v1 Image and Video Processing

Abstract

Synthetic Aperture Radar (SAR) images are usually degraded by a multiplicative noise known as speckle which makes processing and interpretation of SAR images difficult. In this paper, we introduce a transformer-based network for SAR image despeckling. The proposed despeckling network comprises of a transformer-based encoder which allows the network to learn global dependencies between different image regions - aiding in better despeckling. The network is trained end-to-end with synthetically generated speckled images using a composite loss function. Experiments show that the proposed method achieves significant improvements over traditional and convolutional neural network-based despeckling methods on both synthetic and real SAR images.

Keywords

Cite

@article{arxiv.2201.09355,
  title  = {Transformer-based SAR Image Despeckling},
  author = {Malsha V. Perera and Wele Gedara Chaminda Bandara and Jeya Maria Jose Valanarasu and Vishal M. Patel},
  journal= {arXiv preprint arXiv:2201.09355},
  year   = {2022}
}

Comments

Submitted to International Geoscience and Remote Sensing Symposium (IGARSS), 2022. Our code is available at https://github.com/malshaV/sar_transformer

R2 v1 2026-06-24T08:59:20.358Z