gabor filter visualization

《gabor filter visualization》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、GaborFilterVisualizationV.ShivNagaPrasad∗JustinDomkeUniversityofMarylandUniversityofMarylandshivnaga@cs.umd.edudomke@cs.umd.eduABSTRACTthesinusoid.Figure1(c)showsanothersinusoidwhosefre-Wepresentasystemforthevisualizationofanimportantquencyisdoublethatofthepreviousone;Figure1(d)showssi

2、gnalprocessingtechnique-aGaborFilterbank’sresponsethemagnitudeofitsspectrum.Supposeweaddthesetwosi-toanimage.Todothis,onemustovercometheproblemnusoidsthenwewillobtainasignalasshowninFigure1(e).thatnomulti-dimensionalspacecanbeshowninaasingle,Doingaspectralanalysisonthiswouldshowthecomp

3、osi-staticgraph.Weuseaninteractivewidgettochangethetionofthesignal-thetwopeaksinFigure1(f)correspondtovisiblerangeoftheprojecteddimensions,andadditionalthecomponentsinusoids.Fourieranalysishasproventobegraphicswhichsummarizetheresponsesintheprojectedoneofthemostpowerfultoolsinsignalpro

4、cessing.However,dimensions.Thus,thoughweviewthisfour-dimensionalakeyproblemwithFourieranalysisisthatspectralfeaturesspacethrough2-dimensionalprojections,weallowtheuserfromdifferentpartsoftheimagearemixedtogether.Manytounderstandalldimensions,notjusttheplaneofprojec-imageanalysisapplicat

5、ions,e.g.objectrecognition,track-tion.Wefoundthattheimplementedsystemhelpedinget-ing,etc.,requirespatiallylocalizedfeatures.GaborfilterstingabetterunderstandingofGaborfilterresponses.Weareapopulartoolforthistaskofextractingspatiallylocal-thinkthatuseofadomaindependentinteractiontoolandiz

6、edspectralfeatures[1,2].additionalsummarizationgraphicsmaybeusefulinamoregeneralInformationVisualizationsetting.1800.8700.6600.450GeneralTerms0.2GaborFilters,InformationVisualization,HighDimensional040−0.230Data−0.420−0.610−0.8−10Keywords020406080100120140160180200010203040506070809010

7、0(a)(b)GaborFilters,InformationVisualization,HighDimensional1100Data0.8900.6800.4701.INTRODUCTION0.260050Spatialfrequenciesandtheirorientationsareimportantchar-−0.240acteristicsoftexturesinimages.Figure2showsexamples−0.430−0.620ofspatialtextureswithcharacteristicfrequencyandorien-−0.