Salient Object Detection Via Objectness Enhancement

Salient object detection aims to fast segment the object or area which can catch human attention,according to the human visual attention mechanism.At present,salient object detection is regarded as an important pre-processing technology and used in multiple tasks in computer vision field.It can effectively reduce the computational burden of subsequent tasks,thus improve the efficiency of the overall image processing algorithm.It plays a significant role in image processing in the era of big data.In this paper,a salient object detection method via objectness enhancement is proposed.This method consists of three main parts,1)An initial saliency map estimation method based on saliency priors.This model generates an initial saliency map by fusing four prior maps,which can accurately locate salient region.2)An objectness-based salient object detection.First,multiple object proposals are generated by objectness method.Then based on the initial saliency map,a fuzzy set theory is applied to measure the objectness score of the object proposals and integrate them into an objectness map.By integrating the object-level information of the salient object,this model can effectively detect the position and contour of the salient object.3)A graph-based saliency optimization framework.First,the input image is represented as a sparsely connected graph.Then based on the graph model,we build a saliency optimization framework to effectively integrate various prior saliency cues and object-level information.This model can highlight the salient object,suppress the background,as well as maintain the completeness of the salient object.Experimental studies in three benchmark datasets confirmed the superiority of the proposed method over 11 state-of-the-art saliency detection methods in both qualitative and quantitative evaluations.