Abstract
In this study, we address the problem of salient region detection. Recently, salient region detection methods with histogram-based contrast and region-based contrast have given promising results. Center bias is a hypothetical characteristic in human vision system which is applied in many existing salient region detection methods. In this paper, we propose an object-biased Gaussian model to refine the histogram-based contrast method and region contrast method. The proposed algorithm is simple, efficient, and produces full-resolution, high-quality saliency maps. We extensively evaluated our algorithm using traditional salient object detection benchmark, as well as a more challenging co-saliency object detection benchmark. Experimental results demonstrate that the proposed algorithm outperforms the original global contrast methods and other existing salient object detection methods.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call