Blind Spot Network

Blind spot networks (BSNs) are a class of self-supervised image denoising models designed to overcome the challenge of training without paired clean-noisy image data. Current research focuses on improving BSN architectures, particularly by integrating transformers and employing asymmetric masking strategies to enhance performance and address limitations in handling spatially correlated real-world noise. These advancements aim to create more robust and efficient denoising methods for various applications, improving image quality without relying on extensive labeled datasets. The resulting improvements in image restoration have significant implications for diverse fields, including computer vision and medical imaging.