Hidden MRF model-based algorithms for NMR image analysis

Presents a new framework for unsupervised NMR image analysis based on hidden MRF modeling and algorithms. According to the NMR image statistics, two types of hidden MRF models are introduced and justified in terms of stochastic regularization. The image analysis is then formulated as an optimization problem and achieved in two stages: estimate the model parameters to initialize the maximum likelihood solution and conduct finer segmentation through Bayesian decisions using the local context. The solution of the new problem formulation is implemented with an efficient multistage procedure. The experimental results with real NMR images are provided to demonstrate the promise and effectiveness of the proposed technique