资源描述:
《Complex Document Image Segmentation using Localized》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、2004IEEEInternationalConferenceonSystems,ManandCyberneticsComplexDocumentImageSegmentationusingLocalizedHistogramAnalysiswithMulti-LayerMatchingandClusterin6Yen-LinChen,Chung-ChengChiu,andBing-FeiWuDepartmentofElf:ctricalandControlEngineering,NationalChiaoT
2、ungUniversity,1001Ta-HsuehRoad,Hsinchu30050,Taiwan(ylchen,chiu,bwu}@cssp.nctu.edu.twapproachesapplythefeaturesofwaveletcoefficientstoAbstract-Thispaperproposesanewsegmentationextracttext[6]-[7].Theaforementionedmethodsmostlymethodtoseparatethetext?omvarious
3、complexconcentrateonprocessingdocumentimageswhosetextsdocumentimages.Anautomaticmultilevelthresholdingdonotoverlayacomplexbackground.Suchapproachesmethod,basedondiscriminantanalysis,isutilizedtoareeffectiveinextractingcharactersfrommonochromaticrecursivelys
4、egmentaspecijkdblockregionintoseve,mlbackgrounds.However,theycannotsuccessfullysegmentlayeredimagesub-blockr.Thenthemulti-loyerregiin-thetextstringsfromacomplexdocumentimageswhichbasedclusteringmethodispe$ormedtoprocessi,hetextstringsareembeddedwithdecorate
5、dbackgroundlayeredimagesub-blockstoformseveralobjectlayers.objectsortextures.Hencecharacterstringswithdferentilluminations,non-textobjectsandbackgroundcomponentsoresegmentedInthispaper,weproposeaneffectiveregion-basedintoseparateobjectlayers.Afterperformedt
6、iztsegmentationmethodforextractingobjectsfromdifferentextractionprocess,thetextobjectswithdyerentsizes,compleximages.Thecompleximageincludesmanystylesandilluminationsareproperlyextracti?ddifferentobjectssuchascharacters,figures,photographs,Experimentalresul
7、tsontheextractionoftextstringsfiomandcomplexbackgrounds.Thoseobjectsmaybecomplexdocumentimagesdemonstratetheeflectivencssoverlappedbyeachothers.Theproposedmethodutilizesoftheproposedregion-basedsegmentationmethod.twostagestosegmentobjectsfromvariouscompound
8、documentimages.ThefmtstageisautomaticmultilevelKeywnrds:Imagesegmentation,multilevelthresholding,thresholdingbasedonlocalizedhistogramanalysis,whichregion-basedsegmentation,documentanalysis.decomposesd