39 9 2014 9 GeomaticsandInformationScienceofWuhanUniversity Vol.39No.9 Sept.2014 DOI10.13203/j.whugis20130289 1671-8860(2014)09-1103-06 CCD 1 2 3 1 1 1 430079 2 430072 3 100083 CCD (bilateraltotalvariationbtv) CCD ;CCD ; P237.3;TP751 A [412] (Chang E-1) CCD 3 120 m [1] [413] / [2-7] CCD ; / [3] [14] 1(a) 1 57 2 88 ; 1(b) 1 51 2 [15] 95 ; 1(c) ( ) ; 1(d) [16] (IBP) [8] (ML MAP [9] ) 1(c) (POCS) [10] [11] 1(d) 2014-07-03 (4117145061001187); 973 (2011CB707100-6); (6081002) E-mailwlchen 85@163.com
1104 2014 9 1(c) 11.3 L 2N 2 L 1 L 2 1(d) 2.9 P P 1 +λ l=-p m=-p ^Xn -S l xs m y^x {y1y2 ykn 1 N 2 } K N 1 N 2 yk = D kh kf kx +V k k =1 K (1) yk k ;D k ;H k ;F k ( );V k D k=dh k=h V k=v yk = DHF kx +Vk =1 K (2) 1.2 1 ^X ; p L CCD p ;f( X) ;λ Fig.1 Chang E-1CCD Multi-viewImages (3) [4] BTV BTV [17] P P [18] fbtv ( X) = ω m + l X-S l xs yx 1 m l=-p m=-p (4) [19-23] S l x S m y l m ;P 1 ; ω 0<ω<1 (4) [4] (bilat- (3) p L 1 eraltotalvariationbtv) K ^X =argmin X DHF kx -yk 1 + k=1 P P ; λ ω m + l X-S l xs yx 1](5) m l=-p m=-p [4] (5) ; K CCD ^Xn+1 =^X n -β { F kh T D T sign ( DHF k^xn -y k) β ;S x -l 1.1 S -m y -l - m ;D T H T Fk T D H F k ( ) 4 ;I X L1 N 1 L 2 N 2 L 1N 1 [4] X K k=1 ^X =argmin [ DHF kx -yk p p +λf( X) ] k=1 ω m + l [ I-S -m y Sx -l ] sign (3) ( n )} (6)
39 9 CCD 1105 1.3 W kn ^Xn Y k W kn (6) D L 1 L 2 ; H F k ;D T H T D H ; F k Fk T ^X ^X0 1.3.1 {y1y2 yk} ^X ^X0 {y1y2 yk } L 1N 1 L 2N 2 {Y 1 Y 2 Y K } {Y 1 Y 2 Y K } Y i Y i {Y 1 Y i-1 Y i+1 Y K } {W 1 W i-1 W i+1 W K } {Y 1 Y i-1 Y i+1 Y K } Y i {Y 1 Y i-1 Y i+1 Y K } {Y 1 Y i-1 Y i Y i+1 Y K } ^X ^X0 ^X0 2 F k F k T F T k 1.3.3 Brox F k F k T ^X0 Brox [16] 2 CCD 2.1 CCD 3(a) 3(b) 3(c) 113.5 108.7 100.83(d) 3(e) 3(f) 101.8 102.0 100.4 2 Fig.2 EstimationofInitialHighResolution ImageBasedonOpticalFlow 1.3.2 (6) F k ^X0 yk Y k Y k ^X0 W k0 ( 0 ) W k0 ^X0 Y k F k F k T Y k ^X0 Y k W k0 ^X0 Y k ^Xn 3 CCD Fig.3 RadiationCorrectionofChang E-1 CCD Multi-viewImages
1106 2014 9 2.2 CCD 3 40% 6 ( 3 ) CCD 5 2 (5) H 2 5 5 BTV ω=0.6p=2 λ= 0.005 F k F k T 5 BTV 4 3(a) 1 4(a) 4(b) 4(c) 4(d) 4(e) 4(f) 4(b) 4(e) 4 Fig.4 Super-resolutionofChang E-1 CCD Multi-viewImages(Case1) CCD 4(c) 4(f) 5 3(a) 2 5(b) 5(e) 5(a) 5(c) (averagegradi- Fig.5 Super-resolutionofChang E-1 entag) (informationentropyie) 6 CCD ( 1 ) 1 4 5 CCD AG AG CCD CCD 5 3 CCD ( 2 ) CCD Multi-viewImages(Case2)
39 9 CCD 1107 Preserving[J].GeomaticsandInformation Science of Wuhan University201136(5)548-551 (. [J]. 201136(5) 548-551) [6] LiLichunYuQifengYuanYunetal.Super-reso- lution Reconstructionand Higher-Degree Function Deformation ModelBased MatchforChang E-1Lu- narimages[j].scichina Ser E-Tech Sci201040 (3)247-254 (. 6 Fig.6 StructureAnalysis(Case2) [J]. 201040(3)247-254) [7] WeiShiyanShen ZhenrongZhang Shuoet al. MoonRoverImageSuper-resolutionReconstruction Algorithm [J].GeomaticsandInformation Science of Wuhan University201338(4)436-439 (. [J]. 201338(4) 436-439) [8] Irani MPelegS.Motion 1 AnalysisforImageEn- hancementresolutionocclusionandtransparency Tab.1 AssessmentofReconstructedImages [J].Journalof VisualCommunicationsandImage withagandie Representation19934(4)324-335 [9] Hardie R CBarnard K JArmstrong E Eetal. JointMAP Registrationand High-resolutionImage 4-2.81 1.88 3.97 AG EstimationUsingaSequenceofUndersampledIma- 5-3.64 1.80 4.53 ges[j].ieee TransactionsonImage Procesing 4 7.25 7.25 7.19 7.30 IE 199712(6)1621-1633 5 6.11 6.11 5.70 6.23 [10]ErenPESezan MITekalpA M.RobustObject- based High-ResolutionImage Reconstructionfrom Low-resolution Video [J].IEEE Transactionson Image Procesing19976(10)1446-1451 [11]Elad MFeuerA.RestorationofaSingleSuper-res- [1] ZhaoBaochangYangJianfengWenDeshengetal. olutionimagefrom SeveralBlurredNoisyand Un- Designand On-orbit Measurementof Chang E-1 dersampled MeasuredImages [J].IEEE Transac- SateliteCCDStereoCamera[J].SpacecraftEngi- tionsonimage Procesing19976(12)1646-1658 neering200918(1)30-36 ( [12]Ng M KShen HuanfengLam E Yetal.A Total. CCD VariationRegularizationBasedSuper-resolutionRe- [J]. 200918(1)30-36) construction Algorithm for Digital Video [J]. [2] HuangTSTsaiR Y.Multi-frameImageRestora- EURASIP Journalon Advancesin SignalProces- tionand Registration [J].Advancesin Computer visionandimage Procesing1984(1)317-339 ing2007(1)1-16 [13]Elad MHel-OrY.A FastSuper-resolutionRecon- [3] BormanSStevenson R L.Super-resolutionfrom ImageSequencesA Review [C].MidwestSympo- siumoncircuitsandsystemsnotredamein1998 struction AlgorithmforPure Translational Motion and Common Space Invariant Blur [J].IEEE Transactionson Image Procesing200110(8) [4] FarsiuSRobinson M DElad Metal.Fastand Robust Multi-frame Super-resolution [J].IEEE TransactionsonImage Procesing200413(10) 1327-1344 [5] GuoLinChen Qinghu.AdaptiveSuper-resolution ReconstructionofImage Sequence with Structure 1187-1193 [14]BarronJLFleetDJBeaucheminSSetal.Per- formanceofopticalflow Techniques[J].Interna- tionaljournalof Computer Vision199412(1)43-77 [15] OdobezJ MBouthemy P.Roust Multiresolution
1108 2014 9 EstimationofParametricMotion Models[J].Jour- nalof VisualCommunicationandImage Represen- tation19956(4)348-365 [16]BroxTBruhnAPapenbergNetal.HighAccura- cyopticalflow EstimationBasedonaTheoryfor Warping [C].European Conferenceon Computer VisionPragueCzechRepublic2004 [17]Elad M.Super-resolution ReconstructionofImage Sequences-AdaptiveFiltering Approach [D].Isra- elthe Technion-IsraelInstitute of Technology 1996 [18]BakerSKanadeT.Super-resolution OpticalFlow [R].RoboticsInstituteCarnegie Melon Universi- typitsburgh1999 [19] Fransens RStercha CGoolL V.OpticalFlow Based Super-resolutionA Probabilistic Approach [J].Computer VisionandImage Understanding 2007106(1)106-115 [20]MitzelDPockTSchoenemannTetal.VideoSu- per-resolution Using DualityBased TV-L1 Optical Flow [C].DAGM SymposiumonPaternRecong- nitionjenagermany2009 [21]SanguansatPThakulsukanantKPatanavijitV.A Robust Video Super-resolution Using a Recursive LeclercBayesianApproachwithAnOFOF (Optical kingandapplicationsworkshopsfukuoka2012 [22]Krylov ANasonov A.FastSuper-resolutionfrom Video Data Using OpticalFlow Estimation [C]. ingbeijing2008 pixelregistration Approachofthe ADS40Images Basedon OpticalFlow [J].Scienceof Surveying and Mapping200833(6)13-15(. ADS40 [J]. 200833(6)13-15) OpticalFlowBasedSuper-resolutionofChang E-1CCD Multi-viewImages CHEN Wangli 1 SUN Tao 2 CHEN Zhe 3 MA Guorui 1 QIN Qianqin 1 1 StateKeyLaboratoryofInformationEngineeringinSurveyingMappingandRemote SensingWuhanUniversityWuhan430079China 2 SchoolofElectronicInformationWuhanUniversityWuhan430072China 3 ChinaElectronicsTechnologyGroupCorporationBeijing100083China Flow Observation Model)[C].The26thInterna- tionalconferenceonadvancedinformationnetwor- The9thInternationalConferenceonSignalProcess- [23]FanChongGongJianyaZhuJianjunetal.ASub- AbstractAsthedeformationamongmulti-viewimagesfromtheChang E-1CCDisnon-rigidgiven thenon-planarityofthemoonsurfacethetraditionalmethodsbasedontheafinetransform modelare notsuitableforsuper-resolutioncorrection.forthisreasonanewopticalflowbasedsuper-resolution frameworkisproposed.intheframeworkthebtv methodisemployedtoestimateahighresolution imageiteratively.theinitialvaluesforahighresolutionimagearegeneratedbyalmulti-viewimages thatareregisteredbyopticalflowwhiletheopticalflowfieldsareupdatedeachiteration.experimen- talresultsshowthattheproposedmethodperformsbeterforthesuper-resolutionofchang E-1CCD multi-viewimagesthanafinetransformbasedmethods.theresolutionofthereconstructedimageis greatlyimprovedandmoredetailsarevisiblewithoutringing-efects. KeywordsChang E-1;CCD multi-viewimages;super-resolution;opticalflow FirstauthorCHEN WangliPhDcandidatespecializesinrestorationandsuper-resolutionofremotesensingimages.E-mailwlchen 85 @163.com FoundationsupportTheNationalNaturalScienceFoundationofChinaNos.4117145061001187;theNational973ProgramofChina No.2011CB707100-6;FundamentalResearchFundsfortheCentralUniversitiesNo.6081002.