BMVC 2004, Kingston, 7th-9th Sept, 2004
Two--level MRF Models for Image Restoration and Segmentation
M.Rivera (Centro de Investigacion en Matematicas, Mexico) and J.C.Gee
(University of Pennsylvania, USA)
We present a new general Bayesian formulation for
and segmenting piecewise smooth images. This
the associated parameters of the classes within an image,
each image pixel and the number of classes. The intensity
by parametric models based on regularized networks.
the regions (or classes) with complex spatial intensity
identifiable group of simple models. Prior information is
of a two-level Markov random field (MRF). The low-level
information required to recover piecewise restorations,
MRF constraints the segmentation. The high-level MRF
process of simple intensity models into classes.