Shadow Remover Image Shadow Removal Based on Illumination Recovering Optimization(word版本)

《Shadow Remover Image Shadow Removal Based on Illumination Recovering Optimization(word版本)》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、ShadowRemover:ImageShadowRemovalBasedonIlluminationRecoveringOptimizationLingZhang,QingZhang,andChunxiaXiao,Member,IEEE,1057-7149©2015IEEE.Personaluseispermitted,butrepublication/redistributionrequiresIEEEpermission.Seehttp://www.ieee.org/publications_standards/publications/rights/i

2、ndex.htmlformoreinformation.Abstract—Inthispaper,wepresentanovelshadowremovalsystemforsinglenaturalimagesaswellascoloraerialimagesusinganilluminationrecoveringoptimizationmethod.Wefirstadaptivelydecomposetheinputimageintooverlappedpatchesaccordingtotheshadowdistribution.Then,bybuild

3、ingthecorrespondencebetweentheshadowpatchandthelitpatchbasedontexturesimilarity,weconstructanoptimizedilluminationrecoveringoperator,whicheffectivelyremovestheshadowsandrecoversthetexturedetailundertheshadowpatches.Basedoncoherentoptimizationprocessingamongtheneighboringpatches,wefi

4、nallyproducehigh-qualityshadow-freeresultswithconsistentillumination.Ourshadowremovalsystemissimpleandeffective,andcanprocessshadowimageswithrichtexturetypesandnonuniformshadows.Theilluminationofshadow-freeresultsisconsistentwiththatofsurroundingenvironment.Wefurtherpresentseveralsh

5、adoweditingapplicationstoillustratetheversatilityoftheproposedmethod.IndexTerms—Shadowdetection,shadowremoval,shadowmatting,patchmatching,aerialimages.I.IntroductionSHADOWSarenaturalphenomena,whichoccurwhenthelightisblocked.Althoughshadowsprovideimportantvisualcuesforobjectshapeper

6、ception,illuminationposition,objectsocclusion,etc.,shadow-freeimagescanhelptoimprovetheperformanceofthetaskssuchasobjectrecognition,objecttrackingandinformationenhancement[1],[2].Forexample,forhighspatialresolutionremotesenseimage,shadowremovalisverycriticalfortargetidentificationa

7、ndinformationrecovering.Shadowremovalandeditingcanalsoimprovethevisualrealismandphysicalrealisminimageprocessing[3],[4].Shadowremovalisnowanpopularresearchdirectionincomputervisionandimageprocessingcommunities.Manyshadowremovalapproacheshavebeenproposedinthelastdecade(see[2],[5],and

8、[6]forasurvey).Asim