BAYESIAN CHAN-VESE SEGMENTATION FOR IRIS SEGMENTATION

In this paper, we propose a new model as an improvement of active contours without edges model by Chan-Vese to perform iris segmentation. Our proposed algorithm formulates the energy function defined by Chan-Vese as a Bayesian optimization problem. The prior probability is incorporated into the energy function; the prior information of the curve can be integrated with current information provided by likelihood calculation. In order to obtain the desired curve, Maximum a Posteriori (MAP) probability is minimized. Experimental results show that our proposed model gives a more robust performance in iris segmentation compared to the original Chan-Vese model.