Key points are not available for this paper at this time.
Saliency detection, finding the most important parts of an image, has become increasingly popular in computer vision. Existing proposal methods are mostly based on color information, which may not be effective for cluttered backgrounds. We propose a new algorithm leveraging stereopsis to generate optical flow which can obtain addition cue (depth cue) to get the final saliency map. The proposed framework consists of three pathways. The first pathway eliminates the background based on cellular automata. The second pathway gets the optical flow and color flow saliency map. The third pathway calculates a coarse saliency map. Finally, we fuse these three pathways to generate the final saliency map. Besides, we construct a new high-quality dataset with the complex scene to make computer challenge human vision. Experimental results on our dataset and another three popular datasets demonstrate that our method is superior to the existing methods in terms of robustness.
Li et al. (Sun,) studied this question.