资源描述：
《Distance sets for shape filters and shape recognition .pdf》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、1274IEEETRANSACTIONSONIMAGEPROCESSING,VOL.12,NO.10,OCTOBER2003DistanceSetsforShapeFiltersandShapeRecognitionCosminGrigorescu,StudentMember,IEEE,andNicolaiPetkovAbstract—Weintroduceanovelrichlocaldescriptorofanimagepoint,wecallthe(labeled)distanceset,whichisdeterminedby
2、thespatialarrangementofimagefeaturesaroundthatpoint.Wede-scribeatwo-dimensional(2-D)visualobjectbythesetof(labeled)distancesetsassociatedwiththefeaturepointsofthatobject.Basedonadissimilaritymeasurebetween(labeled)distancesetsandaFig.1.(a)Twofamiliarobjects,(b)theircon
3、tours,and(c)partsoftheirdissimilaritymeasurebetweensetsof(labeled)distancesets,wecontours.addresstwoproblemsthatareoftenencounteredinobjectrecog-nition:objectsegmentation,forwhichweformulateadistancesetsshapefilter,andshapematching.Theuseoftheshapefilterisil-arefairlyw
4、ellunderstood[2]–[11],andalargenumberofalgo-lustratedonprintedandhandwrittencharacterrecognitionandrithmsforfeatureextractionhavebeendevelopedincomputerdetectionoftrafficsignsincomplexscenes.Theshapecompar-vision(see,e.g.,[12]–[16]),theneurophysiologyoffeaturein-isonpr
5、ocedureisillustratedonhandwrittencharacterclassifica-tegrationforobjectperceptionandrecognitionisstillunknowntion,COIL-20databaseobjectrecognitionandMPEG-7silhouetteandthedevelopmentofcomputervisionalgorithmsforthispur-databaseretrieval.poseisanongoingactivity[17]–[21]
6、.IndexTerms—Characterrecognition,distanceset,imagedata-Inthispaper,weproposetodescribethespatialinterrela-baseretrieval,MPEG-7,objectrecognition,segmentation,shapetionsbetweenperceptuallysignificantpoints,tobecalledfea-descriptor,shapefilter,trafficsignrecognition.ture
7、points,byassociatingwitheachsuchpointadatastructurecontainingthesetofdistancestoacertainnumberofneigh-I.INTRODUCTIONboringfeaturepoints.Asfeaturescanbeofdifferenttypes,weassociatedifferentlabelswithsuchdistancesandcallthere-HEworld,aswevisuallyperceiveit,isfullofinform
8、ationsultingdatastructurethelabeleddistanceset.WeproposetheTthatweeffortlesslyinterpretascolors,lines,edges,con-setof
