基于區(qū)域的圖像檢索關(guān)鍵技術(shù)研究與系統(tǒng)設(shè)計(jì).pdf

1、西安電子科技大學(xué)碩士學(xué)位論文基于區(qū)域的圖像檢索關(guān)鍵技術(shù)研究與系統(tǒng)設(shè)計(jì)姓名：黃麗娜申請學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：劉志鏡20070101Abstract7AbstractOurworldisdominatedbyvisualinformationandatremendousamountofsuchiIlformationisbeingaddeddaybvdayItwouldbeimpossibletocopewiththi

2、sexplosionofvisualdataunlesstheinformationisorganizedsuchthatwecanretrievethemefficientlyandeffectivelyContentbasedimageretrieval(CBIR)ispresentedthenforimagesExistingCBIRsystemsarebasedonlow—levelvisualfeatnressothebott

3、leneckisthegapbetweenthesevisualfeaturesandsemanticconceptsofimageswhichiscalledsemanticgaPRe#on—basedimageretrieval(RBIR)wasrecentlyproposedasanextensionofcontent—basedimageretrieval(CBIR)whichcannarrowdownthegapInRegio

4、nBasedImageRetrieval(RBIR)，notalltherc：百onsareimportantforretrievingsimilarimagesInretrievaltheuserisofteninterestedinperformingaqueryononlyoneorafewrg：畫。璐ratherthanthewholeimage1memethodismoreefficientInthispaperfirstly

5、thedevelopingprocessofRBIRisintroduced，secondlysometechnologyforaregion—basedimageretrievalframeworkarepresentedandthentheresultsofretrievalexperimentsshowitisaef!ficientsystemLastlyshortcomingsandthedirectofstudyinthefu

6、tureareputtedforwardImagesegmentationisthepreconditionofRBIRAnRBIRsystemautomaticallysegmentsimagesintoavariablenumberofregionsusingtheExpectationMaximizationalgorithmoncombinedfeaturesofeachregion，includingcolortexturea

7、ndshapeFeaturedescriptionisthekeytothesystemInthispapercolorhistogram，gaborwaveletandsevenmomentinvariantsasthefeaturesWealsoireprovedtheintegratedregionmatching(IRM)tosolvethematchingoftheregionsAnefficientindexingofdat

8、abaseandthemethodofdecreasingdimensionsoptimizethesystemInthispaperasetoflowerandupperfunctionsareappliedtofiIterirrelevantimagesLocally—LinearembeddingisusedtodecreasethedimensionsoffeaturevectorsInadditionthispaperaiso

9、introducesthedesignofthesystemandtheresultsofexperimentsInordertoevaluatethismethodanimageretrievaIsysternandexperimentsaredesignedbasedonCORELdatabaseforexample，thelonelyfeaturewithintegratedfeaturetheimprovedmatchingar

10、ithmeticwiththeSCANthejreprovedindexingwiththeorderedindexingCBIRwitIlRBIRandsoonTheexperimentscanbeconcludedthatourmethodcanretrieveatthelevelofobjectsandimprovetheretrievalperformanceAlthoughnowthereitisnotasuccessfull

11、ysystemofRBlR，aprototypehasbeendevelopedtodemonstratethefeasibilityandefficiencyofthesystemIntheendofthispaperlconcludetheshortcomingofthesystemandputforwardthedissertationwithfutareresearchdirectionsforexamplefcferencef