简介：Anewmethodofsuper-resolutionimagereconstructionisproposed,whichusesathree-step-trainingerrorbackpropagationneuralnetwork(BPNN)torealizethesuper-resolutionreconstruction(SRR)ofsatelliteimage.ThemethodisbasedonBPNN.First,threegroupslearningsampleswithdifferentresolutionsareobtainedaccordingtoimageobservationmodel,andthenvectormappingsarerespectivelyusedtothosethreegrouplearningsamplestospeeduptheconvergenceofBPNN,atlast,threetimesconsecutivetrainingarecarriedontheBPNN.Trainingsamplesusedineachstepareofhigherresolutionthanthoseusedintheprevioussteps,sotheincreasingweightsstoreagreatamountofinformationforSRR,andnetworkperformanceandgeneralizationabilityareimprovedgreatly.Simulationandgeneralizationtestsarecarriedonthewell-trainedthree-step-trainingNNrespectively,andthereconstructionresultswithhigherresolutionimagesverifytheeffectivenessandvalidityofthismethod.

简介：Inthispaper,elitistreconstructiongeneticalgorithm(ERGA)basedonMarkovrandomfield(MRF)isintroducedforimagesegmentation.Inthisalgorithm,apopulationofpossiblesolutionsismaintainedateverygeneration,andforeachsolutionafitnessvalueiscalculatedaccordingtoafitnessfunction,whichisconstructedbasedontheMRFpotentialfunctionaccordingtoMetropolisfunctionandBayesianframework.Aftertheimprovedselection,crossoverandmutation,anelitistindividualisrestructuredbasedonthestrategyofrestructuringelitist.ThisprocedureisprocessedtoselectthelocationthatdenotesthelargestMRFpotentialfunctionvalueinthesamelocationofallindividuals.Thealgorithmisstoppedwhenthechangeoffitnessfunctionsbetweentwosequentgenerationsislessthanaspecifiedvalue.Experimentsshowthattheperformanceofthehybridalgorithmisbetterthanthatofsometraditionalalgorithms.

简介：无人的天线车辆(UAV)遥远的成像被坏天气影响，并且获得的图象有低对比的劣势，复杂质地并且变模糊。在这份报纸，我们基于多重散布建议一个盲目deconvolution模型空气点传播功能(APSF)评价到恢复遥感图象。根据Narasimhan分析理论，一个新多重散布恢复模型基于改进二色的模型被建立。然后使用L0标准到估计APSF污迹核的坡度和黑暗隧道的稀少的priors，快Fourier变换被用来由过滤的维纳恢复原来的清楚的图象。由与另外的最先进的方法作比较，建议方法能正确地估计污迹核，有效地移开大气的降级现象，保存图象详细信息并且增加优秀评估索引。

简介：Theeffectofcarrierphaseerrorindigitalimagetransmissionsystemisdiscussed.CodeerrorrateandSNRwithcarrierphaseerrorinGausswhitenoisechannelarecalculated,andthetransmissionsystemissimuatedoncomputer.

简介：Inthispaper,weproposeanewmethodthatcombinescollageerrorinfractaldomainandHumomentinvariantsforimageretrievalwithastatisticalmethod-variablebandwidthKernelDensityEstimation(KDE).TheproposedmethodiscalledCHK(KDEofCollageerrorandHumoment)anditistestedontheVistextexturedatabasewith640naturalimages.ExperimentalresultsshowthattheAverageRetrievalRate(ARR)canreachinto78.18%,whichdemonstratesthattheproposedmethodperformsbetterthantheonewithparametersrespectivelyaswellasthecommonlyusedhistogrammethodbothonretrievalrateandretrievaltime.

简介：Theconceptofdualimagereversibledatahiding(DIRDH)isthetechniquethatcanproducetwocamouflageimagesafterembeddingsecretdataintooneoriginalimage.Moreover,notonlycanthesecretdatabeextractedfromtwocamouflageimagesbutalsotheoriginalimagecanberecovered.Toachievehighimagequality,Luetal.'smethodappliedleast-significant-bit(LSB)matchingrevisitedtoDIRDH.Inordertofurtherimprovetheimagequality,theproposedmethodmodifiesLSBmatchingrevisitedrulesandappliesthemtoDIRDH.Accordingtotheexperimentalresults,theimagequalityoftheproposedmethodisbetterthanthatofLuetal.'smethod.

简介：Side-matchvectorquantization(SMVQ)achievesbettercompressionperformancethanvectorquantization(VQ)inimagecodingduetoitsexplorationofthedependenceofadjacentpixels.However,SMVQhasthedisadvantageofrequiringexcessivetimeduringtheprocessofcoding.Therefore,thispaperproposesafastimagecodingalgorithmusingindirect-indexcodebookbasedonSMVQ(IIC-SMVQ)toreducethecodingtime.Twocodebooks,namedindirect-indexcodebook(II-codebook)andentire-statecodebook(ES-codebook),aretrainedandutilized.TheII-codebookistrainedbyusingtheLinde-Buzo-Gray(LBG)algorithmfromside-matchinformation,whiletheES-codebookisgeneratedfromtheclusteredresidualblocksonthebasisoftheII-codebook.Accordingtotherelationshipbetweenthesetwocodebooks,thecodewordintheII-codebookcanberegardedasanindicatortoconstructafastsearchpath,whichguidesinquicklydeterminingthestatecodebookfromtheES-codebooktoencodetheto-be-encodedblock.TheexperimentalresultsconfirmthatthecodingtimeoftheproposedschemeisshorterthanthatofthepreviousSMVQ.

简介：Withtheadvanceofmultimediatechnologyandcommunications,imagesandvideosbecomethemajorstreaminginformationthroughtheInternet.HowtofastretrievedesiredsimilarimagespreciselyfromtheInternetscaleimage/videodatabasesisthemostimportantretrievalcontroltarget.Inthispaper,acloudbasedcontent-basedimageretrieval(CBIR)schemeispresented.Database-categorizingbasedonweighted-invertedindex(DCWII)anddatabasefilteringalgorithm(DFA)isusedtospeedupthefeaturesmatchingprocess.IntheDCWII,theweightsareassignedtodiscretecosinetransform(DCT)coefficientshistogramsandthedatabaseiscategorizedbyweightedfeatures.Inaddition,theDFAfiltersouttheirrelevantimageinthedatabasetoreduceunnecessarycomputationloadingforfeaturesmatching.ExperimentsshowthattheproposedCBIRschemeoutperformspreviousworkintheprecision-recallperformanceandmaintainsmeanaverageprecision(mAP)about0.678inthelarge-scaledatabasecomprisingonemillionimages.Ourschemealsocanreduceabout50%to85%retrievaltimebypre-filteringthedatabase,whichhelpstoimprovetheefficiencyofretrievalsystems.

简介：EstimationprecisionofDisplacedPhaseCenterAlgorithm(DPCA)isaffectedbythenumberofdisplacedphasecenterpairs,thebandwidthoftransmittingsignalandmanyotherfactors.DetailedanalysisismadeonDPCA'sestimationprecision.AnalysisresultsshowthatthedirectionalvectorestimationprecisionofDPCAislow,whichwillproduceaccumulatingerrorswhenphasecen-ters'trackisestimated.Becauseofthisreason,DPCAsuffersfromaccumulatingerrorsseriously.Toovercomethisproblem,amethodcombiningDPCAwithSubApertureImageCorrelation(SAIC)ispresented.Largesyntheticapertureisdividedintosub-apertures.Microerrorsinsub-apertureareestimatedbyDPCAandcompensatedtorawechodata.Bulkerrorsbetweensub-aperturesareesti-matedbySAICandcompensateddirectlytosub-apertureimages.Afterthat,sub-apertureimagesaredirectlyusedtogenerateultimateSASimage.Themethodisappliedtothelake-trialdatasetofa20kHzSASprototypesystem.ResultsshowthemethodcansuccessfullyremovetheaccumulatingerrorandproduceabetterSASimage.

简介：Classifyingthetextureofgranulesin2Dimageshasarousedmanifoldresearchatten-tionforitstechnicalchallengesinimageprocessingareas.ThisletterpresentsanaggregatetextureidentificationapproachbyjointlyusingGrayLevelCo-occurrenceProbability(GLCP)andBPneuralnetworktechniques.First,upto8GLCP-associatedtexturefeatureparametersaredefinedandcomputed,andtheseconsequentparametersnextserveastheinputsfeedingtotheBPneuralnetworktocalculatethesimilaritytoanyofgivenaggregatetexturetype.Afinitenumberofaggregateimagesof3kinds,witheachcontainingspecifictypeofmineralparticles,areputtotheidentificationtest,experimentallyprovingthefeasibilityandrobustnessoftheproposedmethod.