A superpixel-based CRF saliency detection approach

Salient object detection, experienced several decades, has been an active and popular topic in computer vision. Although a large amount of detection algorithms have been proposed, the obtained saliency maps are still not satisfying enough. To this end, we proposed a simple and novel supervised algorithm to detect a pure background saliency map using conditional random fields (CRF) and saliency cues. Most existing CRF approaches set up the probabilistic graphical models with pixel-wise eight neighborhood grid-shaped graph, while our superpixel level graph handling can not only simplify the model but also promote the performance due to the superpixel level two-ring with pseudo-background neighborhood system. It is intuitive and easy to interpret. As a result, the saliency maps generated by the proposed model have relatively pure background regions. Extensive experimental evaluations on six benchmark datasets with pixel-wise ground truths validated the robustness and effectiveness of the proposed saliency model.