Saliency detection based on boundary feature and smoothness energy function

Abstract Saliency detection is a significant mechanism for image processing tasks. In this paper, we proposed an improved two-stage method that was based on image boundary feature and smoothness energy function. Firstly, we use the color contrast and image boundary feature to obtain the initial prior map and boundary based prior map respectively. Then, we combine the initial prior map and boundary based prior map to generate the improved coarse saliency map. Finally, the smoothness energy function is utilized to refine the coarse saliency map. Experiment on two public datasets demonstrates that our method has a good performance than the state-off-the-art methods.