Previous [ 1] [ 2] [ 3] [ 4] [ 5] [ 6] [ 7] [ 8] [ 9] [ 10] [ 11] [ 12] [ 13] [ 14] [ 15] [ 16]


Journal of Information Science and Engineering, Vol. 32 No. 6, pp. 1435-1454 (November 2016)


Salient Object Detection via Structure Extraction and Region Contrast*


QING ZHANG1, JIAJUN LIN2 AND XIAODAN LI1
1College of Computer Science and Information Engineering
Shanghai Institute of Technology
Shanghai, 201418 P.R. China
2Institute of Automation
East China University of Science and Technology
Shanghai, 200237 P.R. China
E-mail: zhangqing@sit.edu.cn

In this paper, we propose a novel salient object detection approach, which aims in suppressing distractions caused by the small scale pattern in the background and foreground. First, we employ a structure extraction algorithm as a pre-processing step to smooth the textures, eliminate high frequency components and retain the image's main structure information. Second, we segment the texture-suppressed image into perceptually homogenous regions. Third, two saliency feature maps are computed and fused according to the color contrast and center prior cues. To better exploit each pixel's color and position information, we refine the fused saliency map. Experiments on two popular benchmark datasets demonstrate that our proposed approach achieves state-of-the-art performance compared with sixteen other state-of-the-art methods in terms of three popular evaluation measures, i.e., Precision and Recall curve, Area Under ROC Curve and F-measure value.

Keywords: saliency object detection, visual attention, saliency, region of interest, structure extraction

Full Text () Retrieve PDF document (201611_02.pdf)

Received March 14, 2014; revised June 20, 2014; accepted September 28, 2014.
Communicated by Jiann-Liang Chen.
* This work was supported by National Science Foundation of China under Grant No. 61401281 and Science Foundation of Shanghai under Grant No. 14ZR1440700.