Background prior-based salient object detection via deep reconstruction residual

Junwei Han, Dingwen Zhang, Xintao Hu, Lei Guo, Jinchang Ren, Feng Wu

Research output: Contribution to journalArticle

240 Citations (Scopus)

Abstract

Detection of salient objects from images is gaining increasing research interest in recent years as it can substantially facilitate a wide range of content-based multimedia applications. Based on the assumption that foreground salient regions are distinctive within a certain context, most conventional approaches rely on a number of hand designed features and their distinctiveness measured using local or global contrast. Although these approaches have shown effective in dealing with simple images, their limited capability may cause difficulties when dealing with more complicated images. This paper proposes a novel framework for saliency detection by first modeling the background and then separating salient objects from the background. We develop stacked denoising autoencoders with deep learning architectures to model the background where latent patterns are explored and more powerful representations of data are learnt in an unsupervised and bottom up manner. Afterwards, we formulate the separation of salient objects from the background as a problem of measuring reconstruction residuals of deep autoencoders. Comprehensive evaluations on three benchmark datasets and comparisons with 9 state-of-the-art algorithms demonstrate the superiority of the proposed work.
LanguageEnglish
Pages1309-1321
Number of pages13
JournalIEEE Transactions on Circuits and Systems for Video Technology
Volume25
Issue number8
Early online date18 Dec 2014
DOIs
Publication statusPublished - Aug 2015

Fingerprint

Object detection
Deep learning

Keywords

  • salient object detection
  • tacked denoising
  • background prior
  • deep reconstruction

Cite this

Han, Junwei ; Zhang, Dingwen ; Hu, Xintao ; Guo, Lei ; Ren, Jinchang ; Wu, Feng. / Background prior-based salient object detection via deep reconstruction residual. In: IEEE Transactions on Circuits and Systems for Video Technology. 2015 ; Vol. 25, No. 8. pp. 1309-1321.
@article{2dfe27aa0ec74781848a47a5bc9e300c,
title = "Background prior-based salient object detection via deep reconstruction residual",
abstract = "Detection of salient objects from images is gaining increasing research interest in recent years as it can substantially facilitate a wide range of content-based multimedia applications. Based on the assumption that foreground salient regions are distinctive within a certain context, most conventional approaches rely on a number of hand designed features and their distinctiveness measured using local or global contrast. Although these approaches have shown effective in dealing with simple images, their limited capability may cause difficulties when dealing with more complicated images. This paper proposes a novel framework for saliency detection by first modeling the background and then separating salient objects from the background. We develop stacked denoising autoencoders with deep learning architectures to model the background where latent patterns are explored and more powerful representations of data are learnt in an unsupervised and bottom up manner. Afterwards, we formulate the separation of salient objects from the background as a problem of measuring reconstruction residuals of deep autoencoders. Comprehensive evaluations on three benchmark datasets and comparisons with 9 state-of-the-art algorithms demonstrate the superiority of the proposed work.",
keywords = "salient object detection, tacked denoising, background prior, deep reconstruction",
author = "Junwei Han and Dingwen Zhang and Xintao Hu and Lei Guo and Jinchang Ren and Feng Wu",
note = "2015 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works.",
year = "2015",
month = "8",
doi = "10.1109/TCSVT.2014.2381471",
language = "English",
volume = "25",
pages = "1309--1321",
journal = "IEEE Transactions on Circuits and Systems for Video Technology",
issn = "1051-8215",
number = "8",

}

Background prior-based salient object detection via deep reconstruction residual. / Han, Junwei; Zhang, Dingwen; Hu, Xintao; Guo, Lei; Ren, Jinchang; Wu, Feng.

In: IEEE Transactions on Circuits and Systems for Video Technology, Vol. 25, No. 8, 08.2015, p. 1309-1321.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Background prior-based salient object detection via deep reconstruction residual

AU - Han, Junwei

AU - Zhang, Dingwen

AU - Hu, Xintao

AU - Guo, Lei

AU - Ren, Jinchang

AU - Wu, Feng

N1 - 2015 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works.

PY - 2015/8

Y1 - 2015/8

N2 - Detection of salient objects from images is gaining increasing research interest in recent years as it can substantially facilitate a wide range of content-based multimedia applications. Based on the assumption that foreground salient regions are distinctive within a certain context, most conventional approaches rely on a number of hand designed features and their distinctiveness measured using local or global contrast. Although these approaches have shown effective in dealing with simple images, their limited capability may cause difficulties when dealing with more complicated images. This paper proposes a novel framework for saliency detection by first modeling the background and then separating salient objects from the background. We develop stacked denoising autoencoders with deep learning architectures to model the background where latent patterns are explored and more powerful representations of data are learnt in an unsupervised and bottom up manner. Afterwards, we formulate the separation of salient objects from the background as a problem of measuring reconstruction residuals of deep autoencoders. Comprehensive evaluations on three benchmark datasets and comparisons with 9 state-of-the-art algorithms demonstrate the superiority of the proposed work.

AB - Detection of salient objects from images is gaining increasing research interest in recent years as it can substantially facilitate a wide range of content-based multimedia applications. Based on the assumption that foreground salient regions are distinctive within a certain context, most conventional approaches rely on a number of hand designed features and their distinctiveness measured using local or global contrast. Although these approaches have shown effective in dealing with simple images, their limited capability may cause difficulties when dealing with more complicated images. This paper proposes a novel framework for saliency detection by first modeling the background and then separating salient objects from the background. We develop stacked denoising autoencoders with deep learning architectures to model the background where latent patterns are explored and more powerful representations of data are learnt in an unsupervised and bottom up manner. Afterwards, we formulate the separation of salient objects from the background as a problem of measuring reconstruction residuals of deep autoencoders. Comprehensive evaluations on three benchmark datasets and comparisons with 9 state-of-the-art algorithms demonstrate the superiority of the proposed work.

KW - salient object detection

KW - tacked denoising

KW - background prior

KW - deep reconstruction

UR - http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6987333&tag=1

U2 - 10.1109/TCSVT.2014.2381471

DO - 10.1109/TCSVT.2014.2381471

M3 - Article

VL - 25

SP - 1309

EP - 1321

JO - IEEE Transactions on Circuits and Systems for Video Technology

T2 - IEEE Transactions on Circuits and Systems for Video Technology

JF - IEEE Transactions on Circuits and Systems for Video Technology

SN - 1051-8215

IS - 8

ER -