Fast iterative computation of Gaussian mixture parameters and optimal image segmentation

Abstract

We present a fast and accurate parameter estimation method for image segmentation using the maximum-likelihood function. The segmentation is based on a parametric model in which the probability density function of the grey levels in the image is assumed to be a mixture of two Gaussian density functions. For more accurate parameter estimation and segmentation, the algorithm is formulated as a compact iterative scheme. In order to reduce the computation time and to make convergence fast, histogram information is combined into the algorithm. Estimates of the initial values are properly selected for fast convergence. In addition, we rnd the optimal threshold values for several different types of mixture density which have one, two or no intersections between two component densities. The performance of the algorithm is evaluated on a set of artificial and real images and compared with those of other algorithms as well.