基于圖像內(nèi)容自動(dòng)標(biāo)注的海量網(wǎng)絡(luò)圖像檢索.pdf

1、浙江大學(xué)碩士學(xué)位論文基于圖像內(nèi)容自動(dòng)標(biāo)注的海量網(wǎng)絡(luò)圖像檢索姓名：歐廬江申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用指導(dǎo)教師：陳剛20070514浙江大學(xué)碩士學(xué)位論文Abs仃actAbstraetAsanimponant面曲硼ationcarrier，imageisanessemialpartofthenetworkofinfo徹ationrcsourcesWimmerapiddevelopmentofdigitaltechnologvandsca

2、llIlingtcctlll0109y，wecanaccessincreasingimagedataontheimemetHowtoideIni白theimagei—’onnationmneedsmorequicklya11daccllrately仃omvaStnetworkimagedataisthecurrenthotspotinimageremevaIareaThcrefore，onthebaSisoftheformersearc

3、haboutsearchengIneandannotatlOnbasedon蚶瞳ageeOntent，tⅪs口aper口utsfbnvardamassiveWWWima留eretrievalfame、加rkbasedonautomatic鋤otationThemassivedataindexofimagere仃ievalcannotbefinjshedinconce刪callvre仃ievalsystemindependently，an

4、dmedistributedsystemisabasicappmachtofesolve也eproblemofdatasizeandsystemscalabilitySo也emassiveWWWimageretrieval丘ameworkisbasedondis仃ibutedsystemThispaperin仃oduces幽edesignaIldpfincipleofadis廿ibutedretrievalsyst鋤Itdescribe

5、skeytecllIl0109iesandimplementsoft11eimporta】1tmodlllesint11esystemdevelopmentTotakeadvantageofdistributedsystem卸dimprovesystemperfonTlance，thegoaIofourimplememsisthesystembetterscaIab“ity，詘ptabiIityandmorefaun卜to【erance

6、Retrieval丘ameworkqueryisbasedonkeywordsthat廿leuserinputsFortIlenet、Ⅳorkimages，meteXtinits、vebsitecontainsrichsem趾ticiⅢIonnationrelatedtomeimagesThercmeval行鋤eworkstarts、vimextractionofsemantickeywords行omtherelatedtextThjs

7、stepissimilartoordinar)，textre伍evalHOwever’sometimesthewebsitetextisinsu銜ciemtopmvidesemaIltickeywordsSowepm口oseancwapproachofima窯earlno詛tionbasedoni玎1agecontentWhen甘Iereisnoe圩毫ctiverelatedtextformeimages，也esystemaIulota

8、tesmeimage謝mthea口proachbaSedonmacllineleanlingalldstatisticalmodelautomaticallyTheannotationwillbeusedaSthekcywordsfort11eimageretrievalOurapproachisbaSedCo—occurcncemodelWeerlhanceautomaticannotationefrectin3、vays：1)Eff

9、色ctiveimagesegnlemationandimageobjectextraction2)Input油ages、villbem印pedtomorethanonecluster’thusobtaillingmorcprecisesemamic鋤otation，3)Heaviert11eimponaIlceoftlleobectsi11theimagecentertExperimentsshowthatthetechn0109yca