Project description
Many real-world applications require the estimation of images from limited and noisy data. The Bayesian framework provides a way to systematically model the effects of such limitations, as well as incorporate prior information about the images.
The result of this modeling step is a posterior distribution that characterizes the resulting uncertainties and can be used to provide images with its corresponding error bars. In most practical applications only the mode or the mean of this distribution is computed, and uncertainties are ignored.
This project consists of 2 parts:
- Adapt existing methods for uncertainty quantification (UQ) and asses their usability in large-scale imaging applications.
- Develop a practical method for UQ in X-ray CT and validate it on real data from our in-house CT scanner.
