University of Oulu

Yingyue Xu, Xiaopeng Hong, Xin Liu, Guoying Zhao. Saliency detection via bi-directional propagation. Journal of Visual Communication and Image Representation, Volume 53, 2018, Pages 113-121, ISSN 1047-3203.

Saliency detection via bi-directional propagation

Saved in:
Author: Xu, Yingyue1; Hong, Xiaopeng1; Liu, Xin1;
Organizations: 1Center for Machine Vision and Signal Analysis, P.O. Box 4500, 90014, University of Oulu, Finland
Format: article
Version: accepted version
Access: open
Online Access: PDF Full Text (PDF, 10.2 MB)
Persistent link:
Language: English
Published: Elsevier, 2018
Publish Date: 2020-03-14


Recent saliency models rely on propagation to compute the saliency map. Previous propagation methods are single directional, where foreground propagation and background propagation are separate (e.g., only foreground propagation, or background propagation after foreground propagation). Different from the previous approaches, we propose a bi-directional propagation model (BIP) for saliency detection. The BIP model propagates from the labeled foreground superpixels and the labeled background superpixels to the unlabeled ones in the same iteration. A difficulty-based rule is adopted to manipulate the prorogation sequence, which considers both the distinctness of the superpixel to its neighboring ones and its connectivity to the labeled sets. The BIP model outperforms fourteen state-of-the-art saliency models on four challenging datasets, and largely enhances the propagation efficiency compared to single directional propagation models.

see all

Series: Journal of visual communication and image representation
ISSN: 1047-3203
ISSN-E: 1095-9076
ISSN-L: 1047-3203
Volume: 53
Pages: 113 - 121
DOI: 10.1016/j.jvcir.2018.02.015
Type of Publication: A1 Journal article – refereed
Field of Science: 113 Computer and information sciences
Funding: We express deep gratitude to the Academy of Finland, Infotech, Tekes Fidipro Program (Grant No. 1849/31/2015), Tekes project (Grant No. 3116/31/2017), and Natural Science Foundation of China under the contract No. 61772419. The authors also wish to acknowledge CSC – IT Center for Science, Finland, for generous computational resources. Xiaopeng Hong is partly supported by the Natural Science Foundation of China under the contract No. 61572205. Xin Liu is partly supported by the Natural Science Foundation of China under the contract No. 61601362. The authors also wish to acknowledge the supports of NVIDIA Corporation with the donation of the Tesla K40 and K80 GPUs used for this research. Besides, we appreciate all the code and results from the corresponding authors, especially the code of TLLT model from Prof. Chen Gong and the code of BSCA model from Prof. Huchuan Lu.
Copyright information: © 2018 Elsevier Inc. All rights reserved. This manuscript version is made available under the CC-BY-NC-ND 4.0 license