To detect visually salient elements of complex natural scenes, computational bottom-up saliency models commonly examine several feature channels such as color and orientation in parallel. They compute a separate feature ...