简介:Analternativemethodisproposedinthisletterfordescribingthearbitraryshapeandsizeforgranulesin2Dimage.Afterimagebinarization,theedgepointsoncontouraredetected,bywhichthecentroidoftheshapeinquestionissoughtusingthemomentcalculation.UsingPrincipalComponentAnalysis(PCA),themajorandminordiametersarecomputed.Basedonthesignaturecurve-fitting,thefirst-orderderivativeistakensoastoseekallthecharacteristicvertices.Byconnectingtheverticesfound,thesimplifiedpolygonisformedandutilizedforshapeandsizedescriptivepurposes.Thedevelopedalgorithmisrunontwogivenrealparticleimages,andtheexecutionresultsindicatethatthecomputedparameterscantechnicallywelldescribetheshapeandsizefortheoriginalparticles,beingabletoprovideaready-to-usedatabaseformachinevisionsystemtoperformrelateddataprocessingtasks.
简介:Theletterproposesathree-layermanageablemediadistributionnetworksystemarchi-tecturecalledMSPnet,whichisbasedonSessionInitiationProtocol[1]andPeertoPeer(SIPP2P)technology.MSPnetperformsapplication-levelstructuredDHTroutingandresourcelocationamongdomainsandunstructuredonesindomain.Exceptformediadistribution,itcanbeusedtosupportavarietyofP2Papplications,includingvideobroadcasting,videoondemand,VoIP,etc.MSPnetiscomposedofthreelayers,namely,thesignalcontrollayer,themanagementlayer,andthemediatransportationlayer.TheMSPnetprototypeconsistsoftheSIPserver,themanagementserver,themediaserver,andthenodeUserAgent(UA).Resultsfromaprototypeexperimentinalarge-scaleInternetenvironmentshowthatMSPnetisfeasible,scalableandmanageable.
简介:AnewcompactopticalFanofiltersuitableforbiologicalsensingisproposed,whichpatternsphotoncrystalinsinglecrystallinesiliconnanomembranes(SiNMs)andtransferringontotransparentglasssubstrates.Theeffectsofairholesizeandsiliconthicknessonthetransmissioncharacteristicsofnewfilterarenumericallyinvestigatedbyusingthree-dimensionalfinite-differencetime-domain(FDTD)technique,thespectralresponseisalsostudiedbyback-fillingbioliquid.Theresultsshowthatthedipwavelengthwillshifttowardlongerwavelengthbyeitherreducingairholeradiusorfillingbio-liquid.Thenumberofdipswillincreasewiththeincreaseofsiliconthickness.Thesizeofproposedfiltercanbelessthan1mm2.
简介:Videoobjectextractionisakeytechnologyincontent-basedvideocoding.AnovelvideoobjectextractingalgorithmbytwoDimensional(2-D)mesh-basedmotionanalysisisproposedinthispaper.Firstly,a2-Dmeshfittingtheoriginalframeimageisobtainedviafeaturedetectionalgorithm.Then,higherorderstatisticsmotionanalysisisappliedonthe2-Dmeshrepresentationtogetaninitialmotiondetectionmask.Afterpost-processing,thefinalsegmentingmaskisquicklyobtained.Andhencethevideoobjectiseffectivelyextracted.Experimentalresultsshowthattheproposedalgorithmcombinesthemeritsofmesh-basedsegmentingalgorithmsandpixel-basedsegmentingalgorithms,andherebyachievessatisfactorysubjectiveandobjectiveperformancewhiledramaticallyincreasingthesegmentingspeed.