微电子技术论文7485845.pdfVIP

IEEETRANSACTIONSONNUCLEARSCIENCE,VOL.63,NO.5,OCTOBER20162567

ComputedTomographySinogramInpainting

WithCompoundPriorModellingBoth

SinogramandImageSparsity

HanmingZhang,LeiLi,LinyuanWang,YanminSun,BinYan,AilongCai,andGuoenHu

Abstract—ThepresenceofmetalobjectsremainsachallengeMARalgorithmscommonlyfallundertwoclasses:iterative

inx-raycomputedtomography(CT)imaging.Sinogramspassingimagereconstructiontechniquesandprojectioncompletion

throughmetals,calledmetaltrace,usuallyprovideuncorrectedmethods.

informationandareconsideredmissinginCTimagerecon-

struction.ThesparsepriorofanimageinsomeappropriateIntheiterativeimagereconstructionapproach,themissing

transformdomains,definedimplicitsparsity,isoftenusedininformationisgraduallyrecoveredbyutilizingstatisticalmod-

sinograminpaintingmethodsformetaltracerecovery.However,elsorpriorknowledge[4]–[9].However,thelimitedangle

conventionalinpaintingmethodsonlyemploytheimplicitsparsityproblemandtheinconsistenciesinpracticaldataacquisi-

ofasinogramandoftenresultinseveralartifactsintherecon-tioncausetheiterativereconstructionresultstoeasilyintro-

structedimages.Inthispaper,weproposeasinograminpainting

modelwithimplicitsparsityexploitationforbothsinogramandduceblackandwhiteactinomorphicartifactsaroundmetal

image.Anewlyaddedregularizationterm,whichminimizesimplants[10-11].Inrecentyears,aclassofhybridmeth