Salient region detection based on compactness with manifold ranking

The salient region detection aims at completely highlighting the entire objects of interest in an image. It is a challenging problem while a hot topic in computer vision, and has been applied to numerous vision problems such as image compression and object detection. In this paper, we propose a bottom-up salient region detection method via considering the spatial variance of color features in the image space with diffusion process. By considering the spatial distribution of color similarity, foreground and background regions are extracted preliminarily. Besides, in order to produce a pixel-accurate saliency map covering the salient objects more uniformly, we propagate the saliency information through the diffusion process. Experimental results on public benchmark databases show that the proposed approach can uniformly cover salient object regions and effectively suppress background regions.