Microscopic Image Segmentation with Two-dimensional Exponential Entropy Based on Hybrid Microcanonical Annealing

Counting cells and following the evolution of the biological layers are important applications in microscopic imagery. In this paper, a microscopic image segmentation method with two-dimensional (2D) exponential entropy based on hybrid microcanonical annealing is proposed. The 2D maximum exponential entropy does not consider only the distribution of the gray-level information but also takes advantage of the spatial information using the 2D-histogram. The problem with that method is its time-consuming computation that is an obstacle in real time applications, for instance. We propose to combine the microcanonical annealing with the Nelder-Mead method, that was proved very efficient for non convex and combinatorial optimization. As the method is deterministic, the reproduction of the result is guaranteed, thus avoiding any randomization of the solution. The experiments on segmenting microscopic images proved that the proposed method can achieve a satisfactory segmentation with a low computation cost.