Extract salient objects from natural images

Salient objects extraction has been a hot research topic recently, and it is very helpful in many other fields of computer vision, such as image recognition, content-based image retrieval, and image compression. In this paper, we propose a novel method to extract the salient objects from natural images. In the proposed algorithm, we partition the image into patches of homogeneous regions, and then compute the region saliency by combining three different kinds of salient maps, and extract the salient objects eventually by thresholding the quantity of the region saliency. Experiments on our database show the proposed algorithm yields high level of satisfaction.