Scene depth information based image saliency detection

Image saliency detection counts much in image processing and computer vision. Basic features like color, texture and contrast, have been adopted in the construction of saliency map. However, given a complex image, high-level features should be taken into consideration. In general, among the objects with different spatial location, humans tend to focus more on the front-most one. Hence in this paper, a novel method which is combined with scene depth information of objects in an image to detect saliency is proposed. Through this method, the front objects can be distinguished from disturbing background. Besides, we do some improvement to the existing method as well. Experimental results demonstrate that our proposed approach is more precise and effective compared with the original ones.