资源描述:
《dilation matrices for nonseparablenew》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、DilationMatricesforNonseparableBidimensionalWaveletsAnaRuedinDepartamentodeComputaci´on,FacultaddeCienciasExactasyNaturales,UniversidaddeBuenosAiresCiudadUniversitaria,Pab.I.CP1428,CiudaddeBuenosAiresanita@dc.uba.arAbstract.Fornonseparablebidimensionalwav
2、elettransforms,thecho-iceofthedilationmatrixisall–important,sinceitgovernsthedownsam-plingandupsamplingsteps,determinesthecosetsthatgivethepositionsofthefilters,anddefinestheelementarysetthatgivesatesselationoftheplane.Weintroducenonseparablebidimensionalwa
3、velets,andgiveformulaefortheanalysisandsynthesisofimages.Weanalyzeseveraldilationmatrices,andshowhowthewavelettransformoperatesvisually.Wealsoshowsomedistorsionsproducedbysomeofthesematrices.Weshowthattherequirementoftheireigenvaluesbeinggreaterthan1inabs
4、olutevalueisnotenoughtoguaranteetheirsuitabilityforimagepro-cessingapplications,anddiscussotherconditions.Keywords:quincunx,nonseparable,dilation,wavelet.1IntroductionInthelast20years,wavelettransformshaveadquiredgreatimportance.Theyhaveprovedefficientinman
5、ydomains,andthenumberoftheirapplicationsiscontinuallyincreasing.Asatoolinimageprocessing,wavelettransformsarebeingappliedtoimagecompression(theyarethebasisofthestandardJPEG2000[1]),imagedenoising,textureanalysisandpatternrecognition.Waveletsfunctionsareba
6、seswhichhaveexcellenttime(orspace)localization,andgoodfrequencylocalization.Inonedimensiontheyaredilations(changesinscaleinpowersof2)andintegertranslationsofonesamefunctionΨ(x)calledwavelet.Representingasignalinthesebases,weobtainadecompositionofthesignal
7、insumsofdetailsofdifferentresolutions,plusacoarseapproximationofthesignal.Thewavelettransformcanbecalculatedbymeansofthefastwavelettrans-form,throughconvolutionsofthesignalwithtwofilters,onelowpassfilterandonehighpassfilter,followedbyadecimationby2,i.e.onlyth
8、eevencoefficientsareretained.Becauseofthepropertiesofthewavelet,thisoperationisreversible.Toprocessanimage,thetraditionalmanneristoapplytwiceaone-dimensionalwavelettransform,firstbyrowsandthenbycolumns.Thisiscalledthes