A Double Markov Random Field Model for Color Image Segmentation

In this paper, color image segmentation problem is cast as a pixel labeling problem in stochastic framework. The observed color image is assumed to be the degraded version of the image pixel label process. RGB color model is employed to model the color. A new Double Markov Random Field (DMRF) model is proposed to model the intraplane label process and also the interplane label process. The pixel labels are estimated using Maximum a Posteriori (MAP) criterion. A hybrid algorithm is proposed to obtain the MAP estimates and the algorithm is found to converge faster than that of Simulated Annealing (SA) algorithm. The performance of the proposed model is found to be superior to that of using Markov Random Field (MRF) model as the intraplane model. The proposed model yielded satisfactory results for different real images.