Multibranch generative adversarial network for noise reduction in SAR imagery

Satellite imaging is becoming increasingly popular due to its ability to acquire data when traditional methods fail. However, such imagery is often affected by noise, including speckle noise caused by microwave wave interference. In this study, we propose a noise reduction solution based on a Generative Adversarial Network (GAN). The designed network features a multi-column generator that effectively reduces noise while preserving image details. Our method significantly outperforms other traditional filtering techniques, delivering superior image quality, particularly in maintaining essential object features in the images.