Multi-resolution Image Fusion in Remote Sensing

255 Pages·2019·7.237 MB·English

by  Joshi, Manjunath V., Upla, Kishor P

Multi-resolution Image Fusion in Remote Sensing
Thetextoffersdiscussionontoolsandtechniquesofmulti-resolutionimagefusion
with necessary mathematical background. It covers important multi-resolution
fusionconceptsalongwithstate-of-the-artmethodsincludingsuper-resolutionand
multistageguidedfilters. Itincludesin-depthanalysisondegradationestimation,
GaborpriorandMarkovRandomField(MRF)prior.Theuseofedge-preserving
filters in multi-resolution fusion is highlighted. The techniques of compressive
sensing and super-resolution are explained for the benefit of readers. Fusion
methodsarepresentedusingspatialandtransformdomainsinadditiontomodel
based fusion approaches. Real life applications and plenty of multi-resolution
imagesareprovidedinthetextforenhancedlearning.
ManjunathV.JoshiisaProfessoratDhirubhaiAmbaniInstituteofInformation
and Communication Technology (DA-IICT), Gandhinagar, India. He has been
involved in active research in the areas of computer vision, machine learning,
and cognitive radio. He has co-authored two books entitled Motion-Free
Super-Resolution and Digital Heritage Reconstruction using Super-Resolution
andInpainting.
KishorP.UplaisanAssistantProfessoratS.V.NationalInstituteofTechnology
(SVNIT), Surat, India. He has 14 years of teaching experience. His areas of
interest include signal and image processing, machine learning, multi-spectral
andhyper-spectralimageanalysis.

Multi-resolution Image
Fusion in Remote Sensing
Manjunath V. Joshi
Kishor P. Upla
UniversityPrintingHouse,CambridgeCB28BS,UnitedKingdom
OneLibertyPlaza,20thFloor,NewYork,NY10006,USA
477WilliamstownRoad,PortMelbourne,VIC3207,Australia
314to321,3rdFloor,PlotNo.3,SplendorForum,JasolaDistrictCentre,NewDelhi110025,India
79AnsonRoad,#06–04/06,Singapore079906
CambridgeUniversityPressispartoftheUniversityofCambridge.
ItfurtherstheUniversity’smissionbydisseminatingknowledgeinthepursuitof
education,learningandresearchatthehighestinternationallevelsofexcellence.
www.cambridge.org
Informationonthistitle: www.cambridge.org/9781108475129
© CambridgeUniversityPress2019
Thispublicationisincopyright. Subjecttostatutoryexception
andtotheprovisionsofrelevantcollectivelicensingagreements,
noreproductionofanypartmaytakeplacewithoutthewritten
permissionofCambridgeUniversityPress.
Firstpublished2019
PrintedinIndia
AcataloguerecordforthispublicationisavailablefromtheBritishLibrary
ISBN978-1-108-47512-9Hardback
CambridgeUniversityPresshasnoresponsibilityforthepersistenceoraccuracy
ofURLsforexternalorthird-partyinternetwebsitesreferredtointhispublication,
anddoesnotguaranteethatanycontentonsuchwebsitesis,orwillremain,
accurateorappropriate.
To
Smita,NidhiandNinad
—MVJ
Dipika,TakshandTithi
—KPU

Contents
ListofFigures xi
ListofTables xv
Preface xvii
Acknowledgments xix
1 Introduction 1
1.1 CharacteristicsofRemotelySensedImagery 3
1.1.1 Multi-spectralimages 8
1.1.2 Panchromaticimage 10
1.1.3 Hyper-spectralimages 12
1.2 LowSpatialResolutionImaging 14
1.3 ImageFusioninRemotelySensedImages 17
1.4 Multi-resolutionImageFusion: AnIll-posedInverseProblem 18
1.5 IndianRemoteSensingSatellites 20
1.6 ApplicationsofImageFusion 22
1.7 Motivation 25
1.8 OrganizationoftheBook 25
viii Contents
2 LiteratureReview 27
2.1 ProjectionSubstitutionBasedTechniques 29
2.2 Multi-resolutionBasedTechniques 34
2.3 ModelBasedFusionApproaches 41
2.4 Hyper-spectralSharpeningMethods 48
2.5 Conclusion 50
3 ImageFusionUsingDifferentEdge-preservingFilters 52
3.1 RelatedWork 53
3.2 FusionUsingMultistageGuidedFilter(MGF) 54
3.2.1 Multistageguidedfilter(MGF) 55
3.2.2 Proposedapproachusingguidedfilter 57
3.3 FusionApproachUsingDifferenceofGaussians(DoGs) 59
3.3.1 DifferenceofGaussians(DoGs) 60
3.3.2 ProposedapproachusingDoGs 61
3.4 ExperimentalIllustrations 63
3.4.1 Experimentations: Ikonos-2dataset 67
3.4.2 Experimentations: Quickbirddataset 71
3.4.3 Experimentations: Worldview-2dataset 74
3.4.4 Computationalcomplexity 78
3.5 Conclusion 78
4 ImageFusion: ModelBasedApproachwithDegradation
Estimation 80
4.1 PreviousWorks 81
4.2 DescriptionoftheProposedApproachUsingBlockSchematic 83
4.3 Background: ContourletTransform(CT) 85
4.4 ContourletTransformBasedInitialApproximation 85
4.5 ForwardModelandDegradationEstimation 88
4.6 MRFPriorModel 91
4.7 MAPEstimationandOptimizationProcess 94
4.7.1 MAPestimation 94
4.7.2 Optimizationprocess 95
Contents ix
4.8 Experimentations 96
4.8.1 Effectofdecimationmatrixcoefficientsonfusion 100
4.8.2 EffectofMRFparameterg onfusion 102
m
4.8.3 Fusionresultsfordegradeddataset: Ikonos-2 103
4.8.4 Fusionresultsfordegradeddataset: Quickbird 112
4.8.5 Fusionresultsfordegradeddataset: Worldview-2 120
4.8.6 Fusionresultsforun-degraded(original)datasets:
Ikonos-2,QuickbirdandWorldview-2 126
4.8.7 Spectraldistortionatedgepixels 133
4.8.8 Computationaltime 137
4.9 Conclusion 139
5 UseofSelf-similarityandGaborPrior 140
5.1 RelatedWork 141
5.2 BlockSchematicoftheProposedMethod 143
5.3 InitialHRApproximation 144
5.4 LRMSImageFormationModelandDegradationMatrix
Estimation 150
5.5 RegularizationUsingGaborandMRFPriors 152
5.5.1 Optimizationprocess 155
5.6 ExperimentalResults 156
5.6.1 Experimentalsetup 158
5.6.2 Experimentalresultsondegradedandun-degraded
Ikonos-2datasets 159
5.6.3 Experimentalresultsondegradedandun-degraded
Quickbirddatasets 164
5.6.4 Experimentalresultsondegradedandun-degraded
Worldview-2datasets 169
5.6.5 ComparisonoffusionresultswithCSandTV
basedapproaches 174
5.6.6 Computationcomplexity 178
5.7 Conclusion 179