Texture Segmentation Using Gaussian Markov Random Field Models

Segmentation of imagery data has received much attention because of its potential application in several disciplines, e.g., medicine, remote-sensing, robot vision etc.. The image segmentation problem can be specified in the following way: let A be the index set for the pixel values for the whole image; we assume that there exists a partition on A, A = A1 ⋃ A2 ⋃…⋃ Aq such that Ai ≠ ϕ and Ai ∩ Aj = ϕ ∀ i ≠ j, where each Ai is homogeneous in some sense. Thus the problem is to find A1, A2,...,Aq. The segmentation problem has been approached using both structural and statistical procedures.