Saliency Map Detection and Segmentation Based on Global Color Contrast

In this paper, a novel saliency map detection and segmentation method is proposed to detect and segment multiple salient objects. The key idea behind this method is to abstract global color contrast feature to obtain preliminary saliency map, which is later improved by the color dispersion and space information. Based on the saliency map, this method is able to segment multiple salient objects using iteratively and dynamically computed threshold. The experimental results on Achanta database verify the effectiveness of this method for saliency detection and segmentation.