Unsupervised medical image analysis by multiscale FNM modeling and MRF relaxation labeling
Y. Wang, Tülay Adalı, T. Lei · 2002
We derive two types of block-wise FNM model for pixel images by incorporating local context. The self-learning is then formulated as an information match problem and solved by first estimating model parameters to initialize ML solution and then conducting finer segmentation through MRF relaxation.