Removing Speckle Noise from Synthetic-Aperture Radar Images Using Artificial Intelligence Techniques

In recent times, Synthetic-Aperture Radar (SAR) images have gained significant importance due to their extensive use in both military and civilian applications. However, these images are often compromised by speckle noise, which severely hampers their interpretability and utility across various domains. This paper explores a range of image denoising techniques, with a particular focus on deep learning-based approaches, while also considering traditional methods. Given the critical nature of the subject, real SAR images have been rigorously analyzed. The study emphasizes that speckle noise in real SAR images is far more detrimental than in simulated scenarios, revealing the limitations of artificial intelligence algorithms when used in isolation for denoising tasks. To address this challenge, the proposed algorithm in this paper first preprocesses the real data using classical methods before applying a neural network for denoising. Both quantitative metrics and visual assessments demonstrate that this hybrid approach not only outperforms standard denoising methods but also preserves critical information in SAR images more effectively.