A Color Saliency Model for Salient Objects Detection in Natural Scenes

摘要

Detection of salient objects is very useful for object recognition, content-based image/video retrieval, scene analysis and image/video compression. In this paper, we propose a color saliency model for salient objects detection in natural scenes. In our color saliency model, different color features are extracted and analyzed. For different color features, two efficient saliency measurements are proposed to compute different saliency maps. And a feature combination strategy is presented to combine multiple saliency maps into one integrated saliency map. After that, a segmentation method is employed to locate salient objects' regions in scenes. Finally, a psychological ranking measurement is proposed for salient objects competition. In this way, we can obtain both salient objects and their rankings in one natural scene to simulate location shift in human visual attention. The experimental results indicate that our model is effective, robust and fast for salient object detection in natural scenes, also simple to implement.