Abstract

In this paper, we propose a salient object detection algorithm that considers both background and foreground cues. It integrates both coarse salient region extraction and a top-down background weight map measure via boundary label propagation into a unified optimization framework to acquire a refined salient map. The coarse saliency map is additionally fused by three prior components: a local contrast map with greater alignment to physiological law, a global focus prior map, and a global color prior map. During the formation of the background weight map, we first construct an affinity matrix and select nodes existing on the border as labels to represent the background. Then we perform a propagation to generate the regional background weight map. Our proposed model was verified on four benchmark datasets, and the experimental results demonstrate that our method has excellent performance.

© 2017 Optical Society of America

Collaborative multicue fusion using the cross-diffusion process for salient object detection

Jin-Gang Yu, Changxin Gao, and Jinwen Tian
J. Opt. Soc. Am. A 33(3) 404-415 (2016)

Toward adaptive fusion of multiple cues for salient region detection

Hong Li, Enhua Wu, and Wen Wu
J. Opt. Soc. Am. A 33(12) 2365-2375 (2016)

Salient object detection fusing global and local information based on nonsubsampled contourlet transform

Dongmei Liu, Faliang Chang, and Chunsheng Liu
J. Opt. Soc. Am. A 33(8) 1430-1441 (2016)

Cauchy graph embedding based diffusion model for salient object detection

Yihua Tan, Yansheng Li, Chen Chen, Jin-gang Yu, and Jinwen Tian
J. Opt. Soc. Am. A 33(5) 887-898 (2016)

Computational model for salient object detection with anisotropy

Di Wu, Xiudong Sun, Yuannan Xu, Yongyuan Jiang, and Chunfeng Hou
Appl. Opt. 51(11) 1742-1748 (2012)