Image Segmentation based on Multi-objective Evolutionary Algorithms

Image segmentation operation seeks to segment digital image in multiple sets being an open challenge due to there is several criteria that should be considered.

This work proposes a Segmentation Multi-objective Evolutionary Algorithm - SMOEA. Given a set of pair images (image to process and image segmented ideally) and a set of basic image digital processing operations, the proposed algorithm calculates a set of non-dominated solutions (Pareto Set) maximizing the sensitivity and specificity, simultaneously. Each SMOEA-solution is a sub-set of basic image digital processing operations selected in the evolutionary process.

The experimental results on different sets of images confirm that: (i) the image segmentation problem is inherently a multi-objective task due to the Pareto set has solutions in trade-off, and (ii) the proposed approach is a promising algorithm when calculating good Pareto sets.