资源描述：
《基于小波变换的非局部图像去噪方法分析》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、AbstractIIIAbstractFortheinfluenceofexternalenvironment,suchasfilmingtechnique,transmissionequipment,digitalimagewillgetnoisepollutionofdifferentdegreeintheactualapplication.Thiswillaffecttheimagevisualeffectsandsubsequentimageprocessingwork.Inrecentyears,imagedenoisinghasoccupiedanimp
2、ortantpositioninthefieldofimageprocessing,Itsfundamentalpurposeistosolvetheimagedegradationproblemduetonoiseinterference.Currently,lotsofimagedenoisingmethodshavebeenproposed,amongthem,theNon-localmeansalgorithmhasattractedmanyscholarsbecauseofitsexcellentimagedenoisingeffect.Basedonth
3、eimagenonlocalsimilarity,TheBM3Dalgorithmintroducestheconceptofcollaborativefilteringtoimproveimagedenoisingeffectfurther,ithasbeenoneofthestate-of-the-artdenoisingmethods.Inthispaper,ourmajorworkscanbesummarizedasthreeparts:(1)Anon-localimagedenoisingmethodbasedonHWDtransformisderived
4、,themethodselectsappropriatetransformforsimilargroups,basedonthecharacteristicsofinformationcontainedinthegrouppatches,thatistosay,smoothhomogeneousgroupsusewavelettransform,whereasotherswithlinesingularityareassignedtoHWDtransform.Thealgorithmmakesfulluseoftheadvantagesofrepresentingd
5、otsingularityofwaveletandlinesingularityofHWD,itcanobtainmoreaccuratetransformcoefficientstoimprovethequalityofimagerestorationwiththenoiseremovalwhilemaintainingimagedetailinformation.(2)Weproposedanimagedenoisingmethodbasedonnon-localwaveletcoefficientsshrinkage.Thealgorithmintegrate
6、stheimagelocalsparsityconstraintandnon-localsparsityconstraintintoaunifiedvariationalframework,namelythedoubleL1modelinwaveletdomain,itassuresthatthecodingcoefficientsofeachlocalpatchbesparseandbeascloseasthecenterofsimilargroups.Thealgorithmcangetmoreaccurateestimatestoreconstructiont
7、heimageandimprovethedenoisingeffect.(3)Finally,onthebasisofpreviousalgorithm,westudiedthreenonlocalwaveletshrinkagemodelsandanalyzedtheaccuracyandperformanceofthesemodelsinestimatingthewaveletcoefficients.ExperimentsresultshaveshownthatthedoubleL1modelhascomparablebehaviorswithnon-lo
