Genetic algorithm for depth images in RGB-D cameras

In this paper, a new method for unsupervised image segmentation that can be applied to RGB-D (red, green, blue - depth) cameras is presented. The method consists in using a genetic algorithm to optimize the homogeneity of the segmented regions of a depth image. It searches for the best gray level ranges for which the segmentation of the image is closer to the ground truth. Experimental results and comparisons to existing algorithms demonstrate how the proposed method works.