Denoising of biomedical image data: Supervised, Self-Supervised and Unsupervised deep learning approaches

Description

Early approaches to learning-based denoising were supervised, i.e. they required pairs of clean and noisy data for training. While theoretically applicable to many types of data and noise, in practice the requirement for paired data severely limited their applicability.

Later, the community developed self-supervised methods that require only unpaired noisy data for training. Such methods rely on assumptions about the nature of the noise that are often (but not always) correct.

Finally, there are unsupervised methods that train deep generative models of the data and noise. Like self-supervised methods, they can be trained on noisy data only, but instead of outputting a single denoised image, these methods allow us to account for uncertainty in the denoising process by producing samples of possible solutions.

The talk will give an overview of these three types of methods, as well as their strengths and limitations.

Period	25 Jul 2023
Held at	Hannover Medical School, Germany
Degree of Recognition	International