摘要
为了解决当目标不在图像中心或者出现在图像周边时,基于中心先验或者背景先验的显著性检测算法往往会产生错误检测的问题,提出使用目标性作为先验信息得到前景显著图,并且利用乘法运算将其与基于背景先验信息计算的显著图相融合,然后进行空间优化得到单尺度下的显著图,最终显著图为多尺度显著图的加权融合.基于公开数据库的实验结果表明：与目前多种前沿算法相比,本文算法具有更优的检测性能,能够凸显整个显著性目标.
Saliency models based on center or background priors are not effective in dealing with the salient objects not appearing at the image center or locating at the image boundaries.In order to solve the above problems,this paper proposed using objectness to calculate the foreground saliency map which was then fused to background saliency map by multiplication.The fused saliency map was further spatially optimized as one-scale saliency map.The final saliency map was the weighted summarization of saliency maps under different scales.Extensive experimental results on several benchmark datasets show that the proposed methods performs favorably against several advanced saliency detection approaches,and highlight the salient objects more effectively.
作者
刘凡
杨赛
杨慧
林宏达
Liu Fan 1 ,Yang Sai 2, Yang Hui2, Lin Hongda2(1 College of Computer and Information, Hohai University, Nanjing 210098, China; 2 School of Electrical Engineering, Nantong University, Nantong 226019, Jiangsu Chin)
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2018年第3期48-51,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
江苏省普通高校自然科学基金资助项目(16KJB520037)
国家自然科学基金资助项目(61602150)
中国博士后科学基金资助项目(2016M600355,2017T100323)
江苏省博士后科研计划资助项目(1601013B)
南通大学大学生创新训练计划资助项目(2017179)
关键词
视觉显著性
显著性目标检测
目标性
背景先验
空间优化
多尺度融合
visual saliency
salient object detection
objectness
background prior
spatial optimization
multi-scale fusion
