Segmentation of Microscopic Images with NSGA-II

Rocio Ochoa-Montiel, Carlos Sánchez-López, Victor Hugo Carbajal-Gómez, Ever Juárez-Guerra


This paper addresses the problem of multiobjective segmentation on microscopic images by using the evolutionary algorithm NSGA-II. Two objective functions are used at the optimization process: Otsu’s inter-class variance and Shannon’s entropy. A set of 71 images of blood cells are used. From this set, three categories of images are generated: with and without preprocessing, and images with Gaussian noise. Experimental results shown that the use of evolutionary multiobjective techniques like NSGA-II, give satisfactory results in the segmentation for more than one category of images.


Segmentation, multiobjective evolutionary optimization, microscopic images

Full Text: PDF (Spanish)