彩色白細(xì)胞圖像的特征選擇與分類識別.pdf

1、東南大學(xué)碩士學(xué)位論文彩色白細(xì)胞圖像的特征選擇與分類識別姓名：周穎穎申請學(xué)位級別：碩士專業(yè)：生物醫(yī)學(xué)工程指導(dǎo)教師：羅立民鮑旭東20060301AbstractDigitalimageprocessingandanalysishasbeenwidelyappliedinmanymedicalareasAdvancedimageprocessingandpatternrecognitiontechnology，whichisusedinthe

2、sumandsortcountingsofleukocytes，isoneoftheimportantmethodsinmedical—aideddiagnosisThetechnologypartlyreplacingthelaborworksandlighteningtheburdenofdoctorsbyeffectivelyreducedsubjectiveinfluence，highlyincreasesthediagnost

3、icefficiencyBasicstudydoneinthispaperisbasedonthemethodsofautomaticclassificationofleukocytesinperipheralbloodsmearbringingforwardasetofbasicalgorithmsincludingpretreatmentformingandselectingfeaturesaswell[15classificati

4、oncriteria。MostworkhasbeenfocusedonthefeatureextractionandselectionSatisfyingresultshavebeenreachedbytheselectedfeamreparametersactingonleukocytesclassificationAccordingtotheeyeballingexperienceofleukocytesclassification

5、combiningwiththeformerworkinfeatureextractionquantificationaldescribedmorphologyparameterssuchasshapeandsizeofcellalgaandbranchnumbersofnIlcleararee)(cractedonthebasisofsingleleukocytecellsegmentationFollowingworkhasbeen

6、doneincolorparametersextraction，mostofwhicharestatisticalcolorcharacteristicsbasedonregioninformationVariousalgorithms，includingthediscretewavelettransformandalleffectivetextureoperatorlocalbinarypattern，areusedintexture

7、parameterextractionbywhichaneffectivefeaturesubsetisformedConsideringthecomplexityofdifferentparameters，astepbystepstrategyisadoptedinfeatureselectionofleukocytestogetoptimalizedfeatureparametersAsubsetefficienttoclassif

8、icationiscomposedofthefeaturesselectedbyGeneAlgorithmwhichisfirstlyusedintexturefeaturesselectionandfollowedbyTabuSearchadoptedinasecondaryselectionwithinthegeneralfeaturevectorsetsAlsothecapabilityofourmethodiscomparedt

9、othatoftherfeatureselectionalgorithmsSeveralneuralnetworkclassifiermodelsareconstructedforleukocytesrecognitionWithasmallquantityofsamplesourproposedmethodsachieveadecentperformanceofleukocytesclassificationKEYWORDS：Patt