pothesisthatremovesredundanciesinthesensoryinput.Abasicprincipleinvisualsystemistosuppressthere-sponsetofrequentlyoccurringfeatures,whileatthesametimekeepssensitivetofeaturesthatdeviatefromthenorm[9].Therefore,onlytheunexpectedsignalscanbedeliveredtolaterstagesofprocessing.
Fromtheperspectiveofinformationtheory,effectivecodingdecomposetheimageinformationH(Image)intotwoparts:
H(Image)=H(Innovation)+H(PriorKnowledge),H(Innovation)denotesthenoveltypart,andH(PriorKnowledge)istheredundantinformationthatshouldbesuppressedbyacodingsystem.Inthe eldofimagestatistics,suchredundanciescorrespondtostatisticalinvariantpropertiesofourenvironment.Thesepropertieshavebeencomprehensivelydiscussedinliteraturepertain-ingtonaturalimagestatistics[4,25,17,18].Nowitiswidelyacceptedthatnaturalimagesarenotrandom,theyobeyhighlypredictabledistributions.
Inthefollowingsections,wewilldemonstrateamethodtoapproximatethe“innovation”partofanimagebyremov-ingthestatisticalredundantcomponents.Thispart,webe-lieveisinherentlyresponsibletothepoppingupofproto-objectsinthepre-attentivestage.
2.1.Logspectrumrepresentation
Oftheinvariantfactorsofnaturalimagestatistics,scaleinvarianceisthemostfamousandmostwidelystudiedproperty[20,17].Thispropertyisalsoknownas1/flaw.ItstatesthattheamplitudeA(f)oftheaveragedFourierspectrumoftheensembleofnaturalimagesobeysadistri-bution:
E{A(f)}∝1/f.(1)Onalog-logscale,theamplitudespectrumoftheensem-bleofnaturalimages,afteraveragingoverorientations,lies
approximatelyonastraightline.
Althoughthelog-logspectrumistheoreticallymaturedandhasbeenwidelyused,itisnotfavoredintheanaly-sisofindividualimagesbecause:(1)thescale-invariancepropertyisnotlikelytobefoundinindividualimages;(2)thesamplingpointsarenotwell-proportioned,thelowfre-quencypartsspansparselyonthelog-logplane,whereasthehighfrequencypartsdrawtogether,sufferingfromnoise[25].
Insteadofthelog-logrepresentation,inthispaper,weadoptthelogspectrumrepresentationL(f)ofLogspectrumcanbeobtainedbyL(f)=logandlogspectrum animage.A(f)representa-
.Thecomparisonbetweenlog-logtionisshowninFig.1.
Thelogspectrumrepresentationhasbeenusedinaseriesofliteraturepertainingtostatisticalscene

analysis
Source Image
y

tiy
tissnneettnnII ggooLL10
Frequency
Log frequency
10
10
Source Image



y
y
ttisnisneetntnII gogoLLFrequency
10
Log frequency

10
10
Figure1.Examplesoflogspectrumandlog-logspectrum.The rstimageistheaverageof2277naturalimages.

Figure2.Examplesoforientationaveragedcurvesoflogspectra.Thesecurvessharesimilarshape.Thelogspectrumiscomputedfromdown-sampledimage.Thesizeofeachlogspectrumis64×64
[22,23,21,11].Inthefollowingsection,wewillexploitthepoweroflogspectruminsaliencydetectiontasks.Ex-
Spatio-temporal saliency detection using phase spectrum of_工学_高等教育_...Zhang. Saliency Detection: A Spectral Residual Approach. Proc. CVPR, 2007....
Spectral Target-Detect... 11人阅读 7页 2.00 Should we consider a ... Approach based on wave... 2人阅读 5页 2.00 The US Response to ...
Saliency Detection: A Spectral Residual Approach_2007_CVPR_SR 51 清华大学电子工程系 智能图文信息处理研究室 Cerner for Intelligent Image and Document ...
Ice-Induced Fatigue Analysis by Spectral Approach for Offshore Jacket Platforms with Ice-Breaki_专业资料。维普资讯 http://www.cqvip.com CiaOenEgneig,V12,...
( 11) : 1254 - 1259. 〔5 〕 XIAODI HOU,LIQING ZHANG. Saliency detection: a spectral residual approach〔C〕. Proceedings of 2007 IEEE Conference on ...
(Spectralresidual approach, SR) – 将前两种进行...SR模型 ? SR(Spectral Residual)模型是由Hou等人...SDSR(Saliency Detection by SelfResemblance)模型由Seo...
Soft Comput DOI 10.1007/s00500-013-1064-0 METHODOLOGIES AND APPLICATION A study in facial regions saliency: a fuzzy measure approach Pawe? Karczmarek ? ...
2014-CVPR-Saliency(Adaptive Partial Differential Equation Learning for Visual ...Saliency detection: A spectral residual approach. In CVPR, 2007. 6 [15] ...
2011_CVPR_ljduan_Visual Saliency Detection by Spatially Weighted Dissimilarity...Zhang. Saliency detection: A spectral residual approach. CVPR, 2007. [16]...
基于空间信息增量的区域对比度RC 特点 特点 速度快 细节精确 产生高质量显著性图像 计算效率较低 38 Saliency Detection: A Spectral Residual Approach 显著性检测:...
热门文档