Salient region detection via simple local and global contrast representation

Salient region detection is important for many computer vision and computer graphics tasks. In this paper we propose a novel salient region detection framework which consists of attended view estimation and color statistics of input image. Four basic steps are involved. First, based on the cue of center-surround difference, a novel local contrast representation is proposed and a block variance map (BVM) is constructed. Second, by simulating human perception, the attention center and attended view of an image are estimated based on BVM. Third, the color saliency is obtained by a simple global contrast representation. Finally, the full-resolution saliency map is built according to the color saliency. We validate our salient region detection method on two distinct public datasets. The experimental results show that our method outperforms most state-of-the-art methods, reducing the mean absolute error by 41.74% and 28.57% compared to the previous best reported results on the MSRA-1000 and CSSD datasets, respectively.