Salient region detection: an integration approach based on image pyramid and region property. Issue 1 (1st February 2015)

• Record Type: Journal Article
• Authors: Kong, Lingfu; Duan, Liangliang; Yang, Wenji; Dou, Yan
• Is Part Of: IET computer vision. Volume 9:Issue 1(2015)
• Page Start: 85
• Page End: 97
• Publication Date: 2015-02-01
• Subjects: object detection -- object tracking -- computer vision -- image colour analysis -- image segmentation; salient region detection -- integration approach -- image pyramid -- region property -- computer vision tasks -- saliency detection -- high-level vision tasks -- object segmentation -- object tracking -- psychological evidence -- image observations -- colour contrast -- global context -- spatially compact colour distribution -- multiscale image abstraction -- saliency analysis approach -- weighted salient image centre -- saliency estimation model -- MSRA-1000 database -- MSRA-5000 database; Computer vision -- Periodicals; Pattern recognition systems -- Periodicals; 006.37
• DOI: 10.1049/iet-cvi.2013.0285
• Languages: English
• ISSNs: 1751-9632

Abstract

Abstract : Salient region detection is important for many computer vision tasks. The saliency detection results may serve as the basis for further high‐level vision tasks like object segmentation and tracking. In this study, the authors propose an integration approach to detect salient region based on three principles from psychological evidence and observations of images, including colour contrast in a global context, spatially compact colour distribution, multi‐scale image abstraction. Based on the above‐mentioned principles, the authors' saliency analysis approach can be formulated in a unified framework. Moreover, they introduce the weighted salient image centre into their saliency estimation model which can boost the performance of saliency detection. They have evaluated the results of their method on the two publicly available databases, including MSRA‐1000 and MSRA‐5000. The experimental results on the datasets demonstrate the effectiveness of the approaches against the other approaches to analyse image saliency.