Entropy based thresholding for color image segmentation

In this paper we present a simple and efficient technique of color image segmentation based on new approach for color level selection among a widely set of color space. We still prove the dependence between segmentation results and the used color level. Our method is based on entropy-based thresholding that is able to separate different objects based on the calculated amount of information contributed by each one. The first part of this manuscript is how to proceed to select the best color components which improve the segmentation results.