Unsupervised texture segmentation with one-step mean shift and boundary Markov random fields

Abstract

This paper presents an unsupervised texture segmentation method with the one-step mean shift algorithm and the boundary Markov random field. For Gaussian mixture models, the one-step mean shift is capable of determining the boundary points which separate neighboring Gaussian component distribution on histograms. The one-step mean shift algorithm is able to provide a coarse image segmentation result based on the image histogram. In order to improve the segmentation result with the constraints of smoothness, the boundary Markov random field is introduced. In the boundary Markov random field, the multilevel logistic distribution (MLL) is employed for the purpose of smoothing regions with its characteristic of region forming, and the boundary information is added to the energy function of the DLL distribution to preserve the discontinuity at boundaries.