A MODEL FOR SEGMENTATION AND ANALYSIS OF NOISY IMAGES

This article proposes a statistical model for image generation that pr ovides automatic segmentation of images into intensity-differentiated regions and facilitates the quantitative assessment of uncertainty ass ociated with identified image features. The model is specified hierarc hically within the Bayesian paradigm. At the lowest level in the hiera rchy, a Gibbs distribution is used to specify a probability distributi on on the space of all possible partitions of the discretized image sc ene. An important feature of this distribution is that the number of p artitioning elements, or image regions, is not specified a priori. At higher levels in the hierarchical specification, random variables repr esenting emission intensities are associated with regions and pixels. Observations are assumed to be generated from exponential family model s centered about these values.