A review of image denoising techniques: From classical filtering to deep learning and diffusion models

Key Points

• The aim is to review advancements in image denoising techniques and their implications for image quality.
• Reviewed techniques from statistical filtering to deep learning and diffusion models.
• Analyzed various noise models like Gaussian and salt-and-pepper noise.
• Compared performance of classical and modern approaches using metrics like PSNR and SSIM.
• Discussed future directions including hybrid models and self-supervised learning.
• Identified strengths and limitations of different denoising techniques.
• Confirmed that modern deep learning methods outperform classical techniques in certain scenarios.
• Highlighted the importance of quality metrics like PSNR and SSIM in evaluating performance.

Abstract