Supervised Texture Segmentation by Maximising Conditional Likelihood

Abstract

Supervised segmentation of piecewise-homogeneous image textures using a modified conditional Gibbs model with multiple pair-wise pixel interactions is considered. The modification takes into account that inter-region interactions are usually different for the training sample and test images. Parameters of the model learned from a given training sample include a characteristic pixel neighbourhood specifying the interaction structure and Gibbs potentials giving quantitative strengths of the pixelwise and pairwise interactions. The segmentation is performed by approaching the maximum conditional likelihood of the desired region map provided that the training and test textures have similar conditional signal statistics for the chosen pixel neighbourhood. Experiments showthat such approach is more efficient for regular textures described by different characteristic long-range interactions than for stochastic textures with overlapping close-range neighbourhoods.