Saliency Map Detection in Video

In various multimedia applications, recently more advancement have achieved great performance in salient object detection in images, but still there is a big challenging problem for detection of salient object in videos. In salient object detection in videos, images getting from camera tracking of an object or motion of an objects in foreground which involving the motion in the background. In this paper, we proposed salient object detection method based on spatiotemporal saliency where we use spatiotemporal difference and coherence of video content. In which basically saliency map for each key-frame we firstly initialized a method depending on spatial difference on color cue and temporal coherence on motion cue. And then we proceed for saliency of another intra and inter frames. Now by combining these two spatiotemporal difference and coherence, we can conveniently detect salient object in videos. The RANSAC algorithm is used for speed improvement. The proposed method is evaluated and the experimental results shows that our method outperform the state-of-art methods.