Salient Object Detection Based on Background Model

Abstract

In this paper, we propose a novel bottom-up method for detecting salient object. First, a coarse saliency map is generated by according to the biggest symmetric surround model and the color distribution-based model. And then, we design a Gauss filter to enhance the coarse saliency map, and preliminarily establish the background model. Next, the salient feature is defined as the color contrast between background region and the rest. Extensive experiments show the performance of the proposed method which performs better in five evaluation indexes.