Salient object detection based on multi-feature graphs and improved manifold ranking

In this paper, a salient object detection model based on multi-feature and modified manifold ranking is proposed. Different from traditional manifold ranking based models, the graphs in the proposed model are constructed by multiple features, and the energy function of manifold ranking is modified to accurately indicate the queries ranking process. Then, the four boundary regions of the image are ranked respectively based on multi-feature graphs with the improved ranking process to get the boundary based salient maps. And the final salient map is generated by integrating the boundary based maps with boundary connectivity prior. Qualitative and quantitative experiments on five public datasets demonstrate that the proposed model performs better than 10 state-of-the-art models under PR curve and Max F-measure measurements and provides robust and balanced results compared with the other models under MAE and AUC measurements.