資源描述：
《基于內(nèi)容的圖象檢索中特征、索引及交互問(wèn)題研究》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、生墮型蘭墊查盔堂熊主堂垡迨塞塑壁墾羞絲塑ABSTRACTCONTENT-BasedImageRetrieval(CBIR)HasBecomeanImportantResearchFieldinMultimediaInformationProcessinginTheseYears．InThisDissertation，SomeImportantProblemsareDiscussed，IncludingImageFEATURES,ImageINDEXING，andHumanComputerINTERACTION．So
2、meAlgorithmsarePresentedandPerformWellfortheAboveProblemsMeanwhile，aCBIRSurveyandaCBIRSystemsListareAlsoPresented．’l。HEImageStatisticsFeatllre$areEmployedWidelysinceCBIRatItsBeginning．IncludingColor,TextureandOtherFeatures，BecauseofTheirRobusmessforRoration，Sh
3、ift，andScaleChange．ColorQuantizationisanImportantProblemwithA11theColorFeatures；Hence．a(chǎn)NovelAlgorithmisDevelopedforAdaptiveColorNon．EquallyQuantization，andItPerformsWellBasedontheExperimentalResults．Meanwhile．AnotherNewFeatureisBuiltBasedonBothColorandTexture．
4、anda3DCo-occurrenceintheHSVColorSpaceOutFIerformstheTraditionalMethodsonOurDatabase．ALTHOUGHtheStatisticsFeaturesareEmployedforSoManyYears．theGap．WhichBetweenTheseLowLevelFeaturesandtheHighLevelConcepts．isStillaSeriousMaaer,andthelmageStructureFeaturesareUsefu
5、lforitSothisDissertationPresentsSeveralTechniquestoSimplifytheJSEGAlgorithmforImageSegmentationtoExtractLargeSemanticRegionsinImageforCBIR．ExperimentsShowtheAdvantageoftheNewAlgorithmovertheTraditionalJSEGAlgorithmforLargeSemanticRegionsExtraction．AnotherNovel
6、CBlRAlgorithmBasedonRunningSub．BlockswithDifrerentSimilarityWeightsisPresented．bySplittingtheEntireImageintoSub．Blocks．Color-LayoutInformationisUsedtoRetfievalImagesUndertheQueryImageFurtherMore，AtierRetrievingtheImageswiththeSameContentLocatedonDi施rentPartsof
7、theSampledImagesFromtheImageDatabase．SomePost．Processings(suchasFaceRecognition)CanBeIncorporatedtoEnhancetheCBIRSystemsforOtherApplications．THISDissertationPresentsaNovelEfficientSemanticImageClassificationAlgorithmforHigh．1evelFeatureIndexingofHigh．dimension
8、ImageFeatures．ExperimentsShowThattheAlgorithmPerforillsWell．BasedonThisTheory,Anothel"GroundtmthisBuilt．a(chǎn)ndtheImagesareCategorizedintoThreeClasses：City,LandscapeandPerson．TheExperi
