STOCHASTIC MODELING AND ESTIMATION OF MULTISPECTRAL IMAGE DATA

Multispectral images consist of multiple channels, each containing dat a acquired from a different band within the frequency spectrum, Since most objects emit or reflect energy over a large spectral bandwidth, t here usually exists a significant correlation between channels, Due to often harsh imaging environments, the acquired data may be degraded b y both blur and noise. Simply applying a monochromatic restoration alg orithm to each frequency band ignores the cross-channel correlation pr esent within a multispectral image, A Gibbs prior is proposed for mult ispectral data modeled as a Markov random field, containing both spati al and spectral cliques, Spatial components of the model use a nonline ar operator to preserve discontinuities within each frequency band, wh ile spectral components incorporate nonstationary cross-channel correl ations, The multispectral model is used in a Bayesian algorithm for th e restoration of color images, in which the resulting nonlinear estima tes are shown to be quantitatively and visually superior to linear est imates generated by multichannel Wiener and least squares restoration.