Saliency Detection Based on Adaptive Boundary

DOI

10.2991/esm-16.2016.2How to use a DOI?

Keywords

Background detection, saliency detection, adaptive boundary, manifold ranking

Abstract

Background detection has been successful in solving many vision tasks. However, there is a challenging problem that they usually fail to face the situation that object and background are similar. In this paper, we present a novel saliency detection method to address this issue. Firstly, we estimate the rough background probability which can characterize the spatial layout of image regions. Secondly, we propose a robust background measure, called adaptive boundary. It is closer to salient object than boundary of image and can help to distinguish the foreground from background more efficiently. Then, we rank the similarity of the image regions using manifold ranking. The saliency value of each image region is defined based on its relevance to the adaptive boundary. Lastly, the saliency map is generated to exploit integrating saliency results of ranking based on adaptive boundary and boundary of image. Experimental results prove that the proposed algorithm outperforms many of the recent state-of-art and classical algorithms on several datasets.

Copyright

© 2016, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Chen Wang
AU  - Yangyu Fan
AU  - Lei Xiong
PY  - 2016/08
DA  - 2016/08
TI  - Saliency Detection Based on Adaptive Boundary
BT  - Proceedings of the 2016 International Conference on Engineering Science and Management
PB  - Atlantis Press
SP  - 5
EP  - 10
SN  - 2352-5401
UR  - https://doi.org/10.2991/esm-16.2016.2
DO  - 10.2991/esm-16.2016.2
ID  - Wang2016/08
ER  -