Self Supervised Denoising

Self-supervised denoising leverages the inherent redundancy in noisy data to train deep learning models for image restoration without requiring paired clean-noisy image datasets. Current research focuses on improving the robustness and efficiency of these methods, particularly addressing spatially correlated noise in real-world images, often employing blind-spot networks, transformers, and variations of Noise2Noise architectures. These advancements are significant because they enable effective denoising in various applications, including medical imaging (e.g., MRI, OCT, CT), microscopy, and acoustic imaging, where obtaining clean ground truth data is challenging or impossible.