A parallel genetic algorithm for cell image segmentation

In this paper, we propose a parallel genetic algorithm for cell image segmentation under severe noise. Our contribution aims at overcoming the drawback of the slow convenence of the traditional genetic algorithm, which was used in our previous work. A priori knowledge about cell shape is incorporated in our method. That is, an elliptical cell contour model is introduced to describe the boundary of the cell. We firstly obtain the gradient image using Canny edge detector; and then use kernelbased dynamic clustering to find out the image points that have a high probability belonging to each cell. Finally a parallel genetic algorithm is used to adjust the parameters of the cell contour model to find a best matching. The segmentation results of noisy human thyroid and small intestine cell images demonstrate that the proposed method is very successful in segmenting images of elliptically shaped cells.