29卷第3期 微电子学与计算机 V01.29No.3
March
2012年3月 MICROEI。ECTRoNICS&C()MPUTER2012
非清晰区域抑制下的显著对象检测方法
郑 阳,刘纯平,柳 恭,王朝晖
(苏州大学计算机科学与技术学院,江苏苏州215006)
存在检测内容缺失和误检的问题.在CA模型的基础上,引入图像清晰度的视觉反差特性,提出非清晰区域抑制下
的图像显著对象检测方法.该方法以离散度作为判断图像中是否存在清晰度差异的标准,并对存在差异的图像进
行抑制.实验结果表明,非清晰区域抑制的CA方法可以在较好的解决大目标检测和复杂背景误检问题,提高了显
著对象检测精度.
关键词:上下文感知;清晰度;显著性;显著对象检测
中图分类号:TP391 文献标识码:A 文章编号:1000一7180(2012)03--0084--05
DetectionBasedonInhibitionofBlur
Saliency Regions
ZHENG Zhacrhui
Yang,LIUChun-ping,LIUGong,WANG
of Scienceand
Computer 215006,China)
(Department Technology,SoochowUniversity,Suzhou
Abstract:Thereare ofcontent andfalsedetectionfor detection
problems missing Context-Aware(CA)Saliency
modelwhen with scale or
detecting onContext—Awaremodel。with
imagelarge objectcomplexbackground.Based
theintroductionofvisualcontrastcharacteristicof detectionmethodbasedoninhibitionof
imageclarity。asaliency
blur was wasmadeasthe tO
regions dispersion criterion whetherthereexist differencein
proposed.The judge clarity
an the wouldbeinhibitedifthereexistthedifference.The resultshowsthat.the
image,andimage experimental
detectionmethodbasedoninhibitionofblur canbettersolvethe ofcontent
saliency and
regions problem missing
falsedetec