Color image segmentation using markov random field models

摘要

In this thesis, the problem of color image segmentation is address in stochastic framework. The problem is formulated as pixel labelling problem. The pixel labels are estimated using maximum a Posteriori (MAP) criterion. The observed image is viewed as the degraded version of the true labels. The degradation process is assumed to be Gaussian process. The image labels are modeled as Markov Random Field (MRF) model and the Ohta (I1,I2,I3) model is used as the color model.