原创力文档- 0
- 0
- 约5.26万字
- 约 10页
- 2025-10-13 发布于北京
- 举报
PSEUDOFOUR-CHANNELIMAGEDENOISINGFORNOISYCFARAWDATA
HirokiAkiyama,MasayukiTanaka,andMasatoshiOkutomi
TokyoInstituteofTechnology
ABSTRACTareexisting.Manyeffectivedenoisingalgorithmsassum-ing
additivewhiteGaussiannoise(AWGN)havealsobeen
Mostdemosaickingalgorithmsonlyfocusonhandlingproposed[15,16].Buttheirsimplecombinationprovides
noise-freeCFArawdata.Inpractice,theCFArawdataarepoorresults.Onereasonisthatthedemosaickingprocess
corruptedbynoise,whichdegradesdemosaickingperfor-changesthestatisticalpropertiesofthenoise.Theremovalof
mance.Full-colorimagequalitystronglydependsonthenon-AWGNisaverychallengingproblem.
performanceofthedemosaicking.Here,weproposeaCFAThesecondapproachisdenoising-before-demosaicking.
rawdatadenoisingalgorithm.Intheproposedalgorithm,theIfwecansufficientlyreducethenoiseoftheCFArawdata
CFArawdataisconvertedtoapseudofour-channelimagebybeforethedemosaickingprocess,wecansimplyapplythede-
rearrangingpixels.Then,thefour-channeldataaretrans-mosaickingalgorithmforthenoise-freeCFArawdatatothe
formedbasedontheprincipalcomponentanalysis(PCA).denoisedCFArawdata.However,mosthigh-performance
Existinghigh-performancegrayimagedenoisingalgorithmisdenoising
文档评论（0）