Development of New Image Fusion Techniques /
M Munawwar Iqbal Ch
- Rawalpindi, MCS (NUST), 2020
- xv, 108 p
The imagefusionprocessaimstocombinethecrucialinformationfrommultipleim- ages obtainedfromdifferentresources.Numerousimagefusiontechniquesarepro- posed toachievebetterfusionaccuracyhowevertheygenerallyyieldlessinformative edge details,introduceartifactsandhaveblurreddetailsinmultifocusimages.More- over,inmedicalimages,resultantfusedimageshavemissingfinedetailsofdifferent tissues, improperfusionofdifferenttissuesandboundariesarealsonotclearlydemar- cated. This workpresents,differentmulti-focusandmulti-modeltechniquestominimize the above-mentionedissues/limitation.Crossbilateralfilterwhichprovidesbetterper- ceptual qualityiscombinedwithnon-subsampledcontourlettransformtomaximize the usefulinformationpresentinmultiplescalesanddirections.Guidedfilterwhose computing timeisindependentoffiltersizeandwellpreservesedgesiscombinedwith discrete wavelettransformformoreinformativeedgedetails.Alphamapalongwith top hattransformextractscommonpropertiesofdifferentimages.Saliencymapshows more meaningfulrepresentationofimagethereforeitisusedalongwithcrossbilat- eral filtertofusemultimodalbrainimagemodalitieswhichhelptocombinethefacts found inmultiplemedicalimagesthathaveexceptionaldistinctivemodalities.Energy of Laplaciananddiscretewavelettransformbasedfusionofmultimodalbrainimages helps toproperlydemarcateboundariesofdifferenttissues.Shiftinvariantdiscrete wavelettransformandsparsefusionisusedformoreinformativeanatomydetailsof medical imageswhichhelpsradiologistsinevaluatingdifferentmodalityimages. Proposed multifocusimagefusionschemesmaintaindifferentimagedetailslikere- duce edgeblurring,introducelessartifactsandpreservesharpedgedetails.Moreover, suggested multimodalimagefusionschemesformedicalimagespreservethefinede- tails ofdifferenttissues,properfusionofdifferenttissuesandclearlydemarcatethe boundaries toobtainhighqualityfusedimage.Higherquantitativeandqualitativeout- comes areobservedintheproposedframeworksforfusionschemesascomparedto other existingschemes. In generalourresearchworkpresentsmulti-focusandmulti-modelimagefusion problems intwomaindomainsi.e.photographyapplications(toovercometheproblem of depthoffieldofcameras)andbiomedicalimaging.Theproposednovelandeffec- tivealgorithmshavebeentestedonrealworlddatatodemonstratetheirperformance improvement.