Graph Laplacian Based Visual Saliency Detection

Abstract

Detection of salient image regions without prior knowledge of their contents remains a challenge task in computer vision. In this paper, we propose a new saliency detection model based on graph Laplacian computation. This new model has two key steps: firstly, we use an image matting Laplacian model for locating the preliminary visual saliency region. Then, an unsupervised feature selection method in CIELab space is used to improve the accuracy of salient object detection. Experimental results show that the new algorithm can achieve better performance than the existing state of the art.