Markov random field image modelling
Michael McGrath · Open University of Cape Town (University of Cape Town) · 2003
This work investigated some of the consequences of using a priori information in image processing using computer tomography (CT) as an example. Prior information is information about the solution that is known apart from measurement data. This information can be represented as a probability distribution. In order to define a probability density distribution in high dimensional problems like those found in image processing it becomes necessary to adopt some form of parametric model for the distribution. Markov random fields (MRFs) provide just such a vehicle for modelling the a priori distribution of labels found in images. In particular, this work investigated the suitability of MRF models for modelling a priori information about the distribution of attenuation coefficients found in CT scans.