欢迎来到天天文库
浏览记录
ID:55974831
大小:727.07 KB
页数:5页
时间:2020-06-03
《一种高斯尺度混合模型的Shearlet域图像去噪.pdf》由会员上传分享,免费在线阅读,更多相关内容在应用文档-天天文库。
1、JournalofNanchangInstituteofTechnologyISSN1674-007609/14PP49—53Voloume33,Number1,February2014一Imaged-enoi·si‘ngusi●ngG一aussl‘anSCa●IemixturemodelinshearletdomainTIANWei,CAOHanwen,XIEKai,SUNJinliang(1.SchoolofInformationEnslneering;2SchoolofScience,Na
2、nchangInstituteofTechnology,Nanchang330099,China)Abstract:AnewmethodwhichusingGaussianscalemixturemodelintheshearletdomainisproposedforimagedenoising.GaussianscalemixturesaretheproductofaGaussianrandomvectorandalindependenthiddenrandomsealarmultiplie
3、r,whichcanmodeltheneighborhoodsofshearletcoeficientsatadjacentpositionsandscales.Underthismodel,theBayesleastsquareestimationisadoptedtoevaluateeachcoeficient.SimulationswithimagescontaminatedbyadditivewhiteGaussiannoisearecarriedouttoshowthattheperf
4、ormanceinshearletdomainsubstantiallysurpassesthatinthewaveletdomain,bothvisualefectandPSNR.Keywords:imagedenoising;Gaussianscalemixture;Bayesleastsquareestimation;shearlettransform1IntroductionTraditionalwaveletmethodshaveenoughabilitytoeficientlyapp
5、roximatesignalscontainingpointwisesingular-ities,buttheydonotperformaswellwithmultidimensionaldata.So,ordinarywavelettransformdoesnotpossessline·singularitiesandedgesfortwo—dimensionalsignalssuchasnaturalimages.GuoKandLabateDputforwardshearlettransfo
6、rmin2007[¨,whichbreaksthelimitationofthewavelettransform.Ontheotherhand,apriorprobabilitymodelforboththenoiseandforuncorruptedimagesisofcentralimpor-tanceintheperformanceofimagedenoising.Inresentyears,modelshavebeendevelopedtoaccountfornon—Gaussi—anb
7、ehaviorsofimagestatistics.VariousprobabilitymodelssuchasthegeneralizedGaussiandistribution.theGaussian8ca1emixturemode1[4-5】andBesselKformmodelE6]wereproposedinliteraturewithbeRercapabilityinmodelingthedistributionofwaveletcoeficients.Inthispaper,wes
8、tudythestatisticalcharacteristicofthesheadetcoeficients,andthenemploythescalemix-turesofGaussiansdistributionaspriormodeltogaintheBayesianleastsquaresestimatorofshearletcoeficients,andfinallyrealizetheremovingofadditivewhiteGaussiannoisefromtheimage.
此文档下载收益归作者所有