Spatiotemporal salient object detection by integrating with objectness

This paper proposes a novel spatiotemporal salient object detection method by integrating saliency and objectness, for videos with complicated motion and complex scenes. The initial salient object detection result is first built upon both saliency map and objectness map. Afterwards, the region size of salient object is adjusted to obtain the frame-wise salient object detection result by iteratively updating the object probability map, which is the combination of saliency map and objectness map. Finally, in order to enhance the temporal coherence, the sequence-level refinement is performed to generate the final salient object detection result. Experimental results on public benchmark datasets demonstrate that the proposed method consistently outperforms the state-of-the-art salient object detection methods.