Unsupervised Image Segmentation using Tabu Search and Hidden Markov Random Field Model

We propose a Tabu search based Expectation Maximization (EM) algorithm for image segmentation in an unsupervised frame work. Hidden Markov Random Field (HMRF) model is used to model the images. The observed image is considered to be a realization of Gaussian Hidden Markov Random Field (GHMRF) model. The segmentation problem is formulated as a pixel labeling problem. The GHMRF model parameters as well as the image labels are assumed to be unknown. This incomplete data problem is solved using the notions of expectation maximization. The expectation step obtains the MAP estimate of the image labels, assuming the availability of parameter estimates. This is achieved by the proposed Tabu Search Algorithm. The estimated image labels are used to obtain the estimates of parameters in the maximization step. Eventually, the EM algorithm converges to the desired labelization. Our algorithm does not require the proper initial estimates of the parameters. Simulation results are presented for three and four class synthetic images.