@MISC\{IMM2003-02430,
    author       = "K. B. Hilger",
    title        = "Statistical Shape Modelling and Markov Random Field Restoration (invited tutorial and exercise)",
    year         = "2003",
    publisher    = "",
    address      = "",
    url          = "http://www.imm.dtu.dk/~kbh/cisp",
    abstract     = "This tutorial focuses on statistical shape analysis using point distribution models (PDM) which is widely used in modelling biological shape variability over a set of annotated training data. Furthermore, Active Shape Models (ASM) and Active Appearance Models (AAM) are based on PDMs and have proven themselves a generic holistic tool in various segmentation and simulation studies. Finding a basis of homologous points is a fundamental issue in PDMs which effects both alignment and decomposition of the training data, and may be aided by Markov Random Field Restoration (MRF) of the correspondence deformation field between shapes. The tutorial demonstrates both generative active shape and appearance models, and {MRF} restoration on {3D} polygonized surfaces.

 {''}Exercise: Spectral-Spatial classification of multivariate images{''} 
From annotated training data this exercise applies spatial image restoration using Markov random field relaxation of a spectral classifier. Keywords: the Ising model, the Potts model, stochastic sampling, discriminant analysis, expectation maximization."
}