Bayesian classification of multivariate image after MAP reconstruction of noisy channels

Presents a supervised Bayesian classifier that makes use of both spectral signatures and spatial interactions after the preprocessing of clean noisy channels. The authors apply the Markov random field model at both preprocessing and classification stages. They perform the optimization using either coordinate descent or iterated conditional mode. The estimation of filter parameters is accomplished by referring to adjacent channels that have higher signal-to-noise ratio.<
>