Salient Object Detection via Video Spatio-Temporal Difference and Coherence

摘要

Recent advances in salient object detection in images have achieved obvious performance in various multimedia applications, but efficient salient object detection in videos is still a challenging problem. In this paper, we propose a novel salient object detection method based on spatio-temporal difference and coherence of video content. Firstly, we initialize the saliency map for each keyframe based on spatial difference on color cue and temporal difference on motion cue. Then, we propagate the saliency among the regions intra and inter frames according to spatio-temporal coherence, in order to generate the saliency maps for other frames and refine the saliency maps of all the frames. By combining spatio-temporal difference and coherence, we can efficiently detect the salient objects in videos. The proposed method is evaluated on the public UVSD data set, and the experimental results show that our method outperforms the state-of-the-art methods by considering both effectiveness and efficiency.