Salient Object Detection with Self-Representation and Weak Correlation of Salient Region

When salient objects are not similar,the saliency values of these objects are not consistent.To address this problem,we proposed a salient object detection method with self-representation and weak correlation of salient region.Firstly,on the basis of the low rank matrix recovery theory,we introduced Laplace regular term for the salient objects to increase the difference between the salient objects and background.Then we minimized F-norm of self-representation coefficients of salient objects,which make salient objects uniformly highlighted.Finally,we adopted the low rank constraint by an adjustable arctangent function to separate the salient objects from background very well.The experimental results on the public salient object detection data sets show that the proposed method can obtain more complete salient objects and more consistent saliency maps.