Abstract
We propose an unsupervised bottom-up saliency detec-tion approach by exploiting novel graph structure and back-ground priors. The input image is represented as an undi-rected graph with superpixels as nodes. Feature vectors areextracted from each node to cover regional color, contrastand texture information. A novel graph model is proposedto effectively capture local and global saliency cues. Toobtain more accurate saliency estimations, we optimize thesaliency map by using a robust background measure. Com-prehensive evaluations on benchmark datasets indicate that our algorithm universally surpasses state-of-the-art unsupervised solutions and performs favorably against supervised approaches.