基于幀間相關(guān)性的視頻去噪.pdf

1、蘇州大學(xué)碩士學(xué)位論文基于幀間相關(guān)性的視頻去噪姓名：胡麗芳申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：電子與通信工程指導(dǎo)教師：石霏20091001VideoDenoisingBasedonInterframeCorrelationsAbstractVideoDe—noisingBasedonInterframeCorrelationsAbstractInthevideosignal，thereisrelevancebetweenframesWecallitf

2、ramecoherenceVideosequencesareoftendistortedbynoiseduringacquisitiontransmissionorstoragehlordertoimprovevideoqualityandcompressionefficiency，tosavebandwidthortoprepareforfollowuptreatment(videocoding，videoanalysis)，vide

3、odenoisingisimportantTraditionallyvideodenoisingmethodprocesseseachvideoframeseparatelywhichcannotobtainsatisfactoryde—noisingresultsThereasonliesinthattheremainingnoiseortheartifactcausedbyde—noisingisdifferentbetweenea

4、chframewhenthecorrelationbetweenframesisnottakenintoaccountTwocategoriesofmethodsCanbeusedtoimprovethede—noisingresultThefirstcategoryusesmotiondetectionandestimationteducetheblurtingartifactThesecondcategorytreatsavideo

5、sequenceasa3DsignalandprocessesitgloballyInthispaperweproposetwovideode—noisingalgorithms，onebasedonthe3Dcomplexwavelettransformandthresholding，andtheotherbasedon2DcomplexwavelettransformandmotionestimationThe3DCWTrepres

6、entsthemotioninformationbetterthantraditionalseparablewavelettransformsandthereforetheframecoherenceisbetterutilizedMotionestimationextractstheframecoherenceexplicitlyandtemporalandspatialfilteringisperformedtoeffectivel

7、yremovenoiseExperimentsshowthattheproposedmethodsoutperformtraditionalmethodsthatprocesseachframeindependentlyThetwomethodshavetheirownadvantagesandshouldbechosenwiselyaccordingtodifferentapplicationrequirementsKeywords：