4、iseandwhiteGaussiannoise.Imagedenoisingisnecessaryinmedicalimageprocessing,andit’sabasisforfurtherprocessing.Traditionalimagedenoisingmethodscanbedividedintospatialdomaindenoisingandfrequencydomaindenoising.Thetraditionalmethodcanremovethenoiseto some extentwhileblurstheimageedges.Duetotheme
6、.Thenon-localmeansfilteringremovesthenoiseandpreservetheedgeswithoutlosingtoomanyfinestructureanddetails.Inthispaper, thenon-localmeansdenoisingalgorithmisappliedto PET imagedenoising.Thedenoisingperformanceofallconsideredmethodsiscomparedinthreewaysintermsofvisualquality,peaksignaltonoisera
7、tio(PSNR)and edgedetail.Throughtheexperimentitwasfoundthat the non-localmeansfilteringmethodcanachievebetterperformancecomparedtomedianfiltering,wienerfilteringandbilateralfilteringinimagedenoising. Atthesametime,itcaneffectivelyprotecttheimageedge