Speckle Suppression Based on Contextual ConvNeXt Network

In order to obtain a better synthetic aperture radar (SAR) image noise suppression effect, a generative adversarial network (GAN) is constructed based on the ConvNeXt architecture with Contextual Transformer structure. The proposed algorithm significantly removes speckle and exhibits certain generalization performance in other noise scenarios. Experimental results indicate that, compared to traditional denoising algorithms or simple deep learning algorithms alone, the denoising effect of our algorithm is superior, yielding clearer and smoother images without introducing other contaminating content. The visual effect is significantly improved.