Low-level and high-level prior learning for visual saliency estimation

Mingli Song, Chun Chen, Senlin Wang, Yezhou Yang

Research output: Contribution to journalArticle

24 Citations (Scopus)

Abstract

Visual saliency estimation is an important issue in multimedia modeling and computer vision, and constitutes a research field that has been studied for decades. Many approaches have been proposed to solve this problem. In this study, we consider the visual attention problem blue with respect to two aspects: low-level prior learning and high-level prior learning. On the one hand, inspired by the concept of chance of happening, the low-level priors, i.e., Color Statistics-based Priors (CSP) and Spatial Correlation-based Priors (SCP), are learned to describe the color distribution and contrast distribution in natural images. On the other hand, the high-level priors, i.e., the relative relationships between objects, are learned to describe the conditional priority between different objects in the images. In particular, we first learn the low-level priors that are statistically based on a large set of natural images. Then, the high-level priors are learned to construct a conditional probability matrix blue that reflects the relative relationship between different objects. Subsequently, a saliency model is presented by integrating the low-level priors, the high-level priors and the Center Bias Prior (CBP), in which the weights that correspond to the low-level priors and the high-level priors are learned based on the eye tracking data set. The experimental results demonstrate that our approach outperforms the existing techniques.

Original languageEnglish (US)
Pages (from-to)573-585
Number of pages13
JournalInformation Sciences
Volume281
DOIs
StatePublished - Oct 10 2014
Externally publishedYes

Fingerprint

Saliency
Color
Computer vision
Visual Attention
Eye Tracking
Statistics
Spatial Correlation
Conditional probability
Large Set
Computer Vision
Multimedia
Experimental Results
Modeling
Demonstrate
Vision
Object
Learning
Relationships
Model

Keywords

  • High-level prior learning
  • Low-level prior learning
  • Visual saliency estimation

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Theoretical Computer Science
  • Software
  • Computer Science Applications
  • Information Systems and Management
  • Artificial Intelligence

Cite this

Low-level and high-level prior learning for visual saliency estimation. / Song, Mingli; Chen, Chun; Wang, Senlin; Yang, Yezhou.

In: Information Sciences, Vol. 281, 10.10.2014, p. 573-585.

Research output: Contribution to journalArticle

Song, Mingli ; Chen, Chun ; Wang, Senlin ; Yang, Yezhou. / Low-level and high-level prior learning for visual saliency estimation. In: Information Sciences. 2014 ; Vol. 281. pp. 573-585.
@article{7dbc1026b1824c4289a2e094739372ad,
title = "Low-level and high-level prior learning for visual saliency estimation",
abstract = "Visual saliency estimation is an important issue in multimedia modeling and computer vision, and constitutes a research field that has been studied for decades. Many approaches have been proposed to solve this problem. In this study, we consider the visual attention problem blue with respect to two aspects: low-level prior learning and high-level prior learning. On the one hand, inspired by the concept of chance of happening, the low-level priors, i.e., Color Statistics-based Priors (CSP) and Spatial Correlation-based Priors (SCP), are learned to describe the color distribution and contrast distribution in natural images. On the other hand, the high-level priors, i.e., the relative relationships between objects, are learned to describe the conditional priority between different objects in the images. In particular, we first learn the low-level priors that are statistically based on a large set of natural images. Then, the high-level priors are learned to construct a conditional probability matrix blue that reflects the relative relationship between different objects. Subsequently, a saliency model is presented by integrating the low-level priors, the high-level priors and the Center Bias Prior (CBP), in which the weights that correspond to the low-level priors and the high-level priors are learned based on the eye tracking data set. The experimental results demonstrate that our approach outperforms the existing techniques.",
keywords = "High-level prior learning, Low-level prior learning, Visual saliency estimation",
author = "Mingli Song and Chun Chen and Senlin Wang and Yezhou Yang",
year = "2014",
month = "10",
day = "10",
doi = "10.1016/j.ins.2013.09.036",
language = "English (US)",
volume = "281",
pages = "573--585",
journal = "Information Sciences",
issn = "0020-0255",
publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Low-level and high-level prior learning for visual saliency estimation

AU - Song, Mingli

AU - Chen, Chun

AU - Wang, Senlin

AU - Yang, Yezhou

PY - 2014/10/10

Y1 - 2014/10/10

N2 - Visual saliency estimation is an important issue in multimedia modeling and computer vision, and constitutes a research field that has been studied for decades. Many approaches have been proposed to solve this problem. In this study, we consider the visual attention problem blue with respect to two aspects: low-level prior learning and high-level prior learning. On the one hand, inspired by the concept of chance of happening, the low-level priors, i.e., Color Statistics-based Priors (CSP) and Spatial Correlation-based Priors (SCP), are learned to describe the color distribution and contrast distribution in natural images. On the other hand, the high-level priors, i.e., the relative relationships between objects, are learned to describe the conditional priority between different objects in the images. In particular, we first learn the low-level priors that are statistically based on a large set of natural images. Then, the high-level priors are learned to construct a conditional probability matrix blue that reflects the relative relationship between different objects. Subsequently, a saliency model is presented by integrating the low-level priors, the high-level priors and the Center Bias Prior (CBP), in which the weights that correspond to the low-level priors and the high-level priors are learned based on the eye tracking data set. The experimental results demonstrate that our approach outperforms the existing techniques.

AB - Visual saliency estimation is an important issue in multimedia modeling and computer vision, and constitutes a research field that has been studied for decades. Many approaches have been proposed to solve this problem. In this study, we consider the visual attention problem blue with respect to two aspects: low-level prior learning and high-level prior learning. On the one hand, inspired by the concept of chance of happening, the low-level priors, i.e., Color Statistics-based Priors (CSP) and Spatial Correlation-based Priors (SCP), are learned to describe the color distribution and contrast distribution in natural images. On the other hand, the high-level priors, i.e., the relative relationships between objects, are learned to describe the conditional priority between different objects in the images. In particular, we first learn the low-level priors that are statistically based on a large set of natural images. Then, the high-level priors are learned to construct a conditional probability matrix blue that reflects the relative relationship between different objects. Subsequently, a saliency model is presented by integrating the low-level priors, the high-level priors and the Center Bias Prior (CBP), in which the weights that correspond to the low-level priors and the high-level priors are learned based on the eye tracking data set. The experimental results demonstrate that our approach outperforms the existing techniques.

KW - High-level prior learning

KW - Low-level prior learning

KW - Visual saliency estimation

UR - http://www.scopus.com/inward/record.url?scp=84904760190&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84904760190&partnerID=8YFLogxK

U2 - 10.1016/j.ins.2013.09.036

DO - 10.1016/j.ins.2013.09.036

M3 - Article

AN - SCOPUS:84904760190

VL - 281

SP - 573

EP - 585

JO - Information Sciences

JF - Information Sciences

SN - 0020-0255

ER -