“…These values define a small cube used to group the different pixels into values taken from a discrete set of color classes. The values for these thresholds that are mostly based in the RGB color space may be set in different ways, including manually [1], using decision trees [2], trained by neural networks [3], and also by dynamically adjusting the initial color calibration using automatic color calibration [4,5,6]. Other approaches, including hie rarchical models, can be found in [7,8,9,10].…”