Object extraction in a quantum inspired soft computing environment

Abstract

Differentkindsofnoiseremovalintelligenttechniquesarebeingusedtoex- newlinetract objectsbymeansofremovalofnoiseartifacts.Amongtheintelligenttools newlinein vogue,theMulti-LayerSelfOrganizingNeuralNetworkarchitecture(ML- newlineSONN) architectureisefficientinextractingobjectsfromblurredandnoisybi- newlinenary images.Grayscaleimagesaresegmentedusingthenetworkbyresorting newlineto amultilevelsigmoidalactivationfunctionsoastoelicitnetworkresponses newlineto allpossiblegrayscales.AparallelMulti-LayerSelfOrganizingNeuralNet- newlinework architecture(PSONN)comprisingthreeparallelMulti-LayerSelfOrganiz- newlineing NeuralNetworkarchitectureswith/withoutmultilevelsigmoidalactivation newlinefunctions canbeusedtoextractpureandtruecolornoisyimages. newlineIn thisthesis,aquantumversionoftheMLSONNarchitectureisproposedtoex- newlinetract andsegmentnoisyimages.TheproposedQMLSONNnetworkarchitecture newlinecomprises threedifferentlayersviz.inputlayer,hiddenlayerandoutputlayer newlineenvisaged by qubit neurons,whereeach qubit neuronineachlayercorresponds newlineto eachpixeloftheinputimage.Theinterconnectionweightsbetweenthediffer- newlineent layersarerepresentedasquantumrotationgates.The qubit neuronhasthe newlineability toextracttheobjectsbyresortingtotheprincipleofquantumsuperposi- newlinetion. newlineThe qubit neuronsofthethreenetworklayersprocesstheimageinformationus- newlineing quantumrotationgatesasinterconnectionweightsguidedbyaquantum newlinebackpropagationalgorithm.Attheoutputlayer,aquantummeasurementis newlineused todestroythequantumstatestoyieldnetworkoutputswhiletheinter- newlineconnection weightsareadjustedusingaquantumbackpropagationalgorithm. newlineA parallelversionoftheproposedarchitecturecomprisingthreeindependent newlineQMLSONN architectures(referredtoasQPSONN)witheacharchitectureen- newlinetrustedforprocessingthethreeindividualprimarycolorcomponents(R,G,B), newlineis alsoproposedtoextractcolorobjectsfrompurecolornoisyimages. newlineA functionalmodificationoftheproposedQMLSONNarchitectureisachieved newlineby incorporatingaMultilevelSigmoidal(MUSIG)activationfunctionsothatthe newlinenetwork isabletosegmentmultilevelgrayscaleimages.Onsimilarlines,the newlineQPSONN architectureisalsomodifiedbyincorporatingmultilevelcharacte

Description

Keywords

Citation

item.page.endorsement

item.page.review

item.page.supplemented

item.page.referenced