水平未知时一种图像恢复正则化算法(图像和数字(3)
发布时间:2021-06-06
发布时间:2021-06-06
壁曼壅王墨塾壁=塑璺堡堡墨垂型垡基鎏
Aregularizationmethodforimagerestorationwithoutknowingnoise
level
ABS’I’RACT
Imagerestorationisoneofthebasicproblemsindigitalimage
processing.Inrecentyears,thetechnologyiswidelyappliedinmang
fields,suchasradioastronomy,sateliteremotesensing,medicalimag-
ingandindustyvision.Theusualmethodsincluderegularization
method,iterativemethod,stochasticmethodandSOon.
Inthisthesis.animprovedcombiningregularizationframework
oftwoparameterstype,proposedbyR.YoumaranandA.Adler,is
presentedanddiscussedforstableimagerestorationwithoutknowing
noiseleveloftheinputdata.Somedetailsoftheimprovementsin—
cludetheadoptionofwell—-knownL-curvecriterionwith
techniqueforaaspeed uproughinitialvalueofregularizationparameter,the
translationofEulerequationrequiringverylargeamoutofstorage
toanequationrequiringmuchlessmemory,therelatedanalysisand
proof,thedesignofthenewalgorithmandSOon.
Thetheoreticanalysisandnumericalexperimentsresultshave
shownthatournewalgorithmhasfastconvergencespeedandgood
arerestorationeffect,andvaluesofISNRimprovedupon0.3dbgen—
onerally.We
inachieveallthenumericalexpermentstogiveapersenalcomputerMatlablanguageandwishhandonthisworks.
KEYWORDS:imagerestoration,regularizationmethod,L—curve
criterion,choosingregularizationparmeter,frameworkoftwoparam—
eterstype
Il