Salient Object Detection Based on the Proto-objects and Background Prior
Gaopeng Zhao, Mingfeng Yin, Liang He · The Open Cybernetics & Systemics Journal · 2015
Image salient object detection is useful for applications like image retrieval, object recognition etc. Aiming to obtain the pixel-level saliency map, most traditional saliency models ignore the object-level information and become inappropriate in complex scenes.A new salient object detection method is proposed in this paper by incorporating the notion of object and background prior directly into the saliency measurement.Firstly the collection of proto-objects is obtained by adapting selective search method.Secondly for each proto-object, we measure its saliency score by computing the distance of histogram feature between the proto-object and the prior background region, then the saliency scores are ranked, and the top-ranked proto-object is considered as the salient object.The publicly MSRA dataset is adapted for experiment evaluation, compared with several other state-of-the-art methods, the proposed method produces superior performance.Experimental results show that the proposed method is effective.