Metaheuristics for unsupervised classification

In this work we have approached the image segmentation problem by means of metaheuristics which have proven to be very efficient in providing good approximate solutions to various optimization problems. The first step was to reformulate the segmentation problem into a single-objective optimization problem in a first step and a multi-objective one in a second step. The second step consisted in the extraction of the spectral and textural information of the image. The extraction of the textural information was done using co-occurrence matrices. An equiprobable requantization has been done beforehand in order to reduce the computation time. For the spectral information we considered the mean and the variance of the gray levels on the image.