文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > IT/计算机 > 计算机软件及应用 > P-Saliency Detection--A Spectral Residual Approach
侵权投诉

P-Saliency Detection--A Spectral Residual Approach

1 image

P-Saliency Detection--A Spectral Residual Approach

P-Saliency Detection--A Spectral Residual Approach

15log intensity

log intensity

5log intensity

100

20406080100120

frequency

frequency

frequency

Figure3.Curvesofaveragedspectraover1,10and100images.

amplesofthelogspectraarepresentedinFig.2.We ndthatthelogspectraofdifferentimagessharesimilartrends,thougheachcontainingstatisticalsingularities.Fig.3showsthecurvesofaveragedspectraover1,10and100images,respectively.Thisresultsuggestsalocallinearityintheav-eragedlogspectrum.

2.2.Fromspectralresidualtosaliencymap

Similaritiesimplyredundancies.Forasystemaimingatminimizingtheredundantvisualinformation,itmustbeawareofthestatisticalsimilaritiesoftheinputstimuli.Therefore,indifferentlogspectrawhereconsiderableshapesimilaritiescanbeobserved,whatdeservesourattentionistheinformationthatjumpsoutofthesmoothcurves.Webelievethatthestatisticalsingularitiesinthespectrummayberesponsibleforanomalousregionsintheimage,whereproto-objectsarepoppedup.

Givenaninputimage,thelogspectrumL(f)iscom-putedfromthedown-sampledimagewithheight(orwidth)equals64px.Theselectionoftheinputsizeisrelatedtovi-sualscale.TherelationshipbetweenvisualscaleandvisualsaliencyisdiscussedinSection3.1.

IftheinformationcontainedintheL(f)isobtainedpre-viously,theinformationrequiredtobeprocessedis:

H(R(f))=HL(f)|A(f),(2)whereA(f)denotesthegeneralshapeoflogspectra,which

isgivenaspriorinformation.R(f)denotesthestatisticalsingularitiesthatisparticulartotheinputimage.Inthispaper,wede neR(f)asthespectralresidualofanimage.ShowninFig.3,theaveragedcurveindicatesalocallin-earity.Therefore,itisreasonabletoadoptalocalaverage lterhn(f)toapproximatetheshapeofA(f).Inourex-periments,nequals3.Changingthesizeofhn(f)alterstheresultonlyslightly(seeFig.5).TheaveragedspectrumA(f)canbeapproximatedbyconvolutingtheinputimage:

A(f)=hn(f) L(f),

wherehn(f)isann×nmatrixde nedby:

1

(3)

hn(f)=

上一页第3页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email