A SAR Image Speckle Suppression Method Based on Transformer and Frequency Features

Synthetic Aperture Radar (SAR) image speckle is a type of image noise caused by the coherence of terrain, which can degrade the resolution and detection capability of SAR images. To address this challenge, this paper proposes a speckle suppression network based on Transformer and frequency features. The network first performs initial denoising on the images, followed by wavelet transformation to retain high-frequency information, and then inputs the results into the network to guide speckle removal. The main body of the network employs Transformer as the encoder for feature extraction, facilitating better learning of the relationships between different regions. Experiments conducted on both synthetic and real datasets validate the effectiveness of the proposed method, showing superior performance compared to existing classical methods.