Visual Saliency Detection Based on Disperse Degree of Color

Abstract

Saliency detection aims to focus attention on the important parts of a map, which is an excellent ability of human visual system. In this paper, we present a saliency detection model based on the principle that the pixels belong to the background are more disperse than the ones of the target area. Color contrast in different channels is employed to classify the pixels. Our method outperformed five art-of-the-state saliency detection algorithms on the ground-truth evaluation experiment by achieving good precision-recall curve and indicates that disperse degree of color is an importance feature in saliency detection.