基于數(shù)字圖像處理的彩色標(biāo)記區(qū)域提取方法的研究.pdf

1、東南大學(xué)碩士學(xué)位論文基于數(shù)字圖像處理的彩色標(biāo)記區(qū)域提取方法的研究姓名：沈建勛申請學(xué)位級別：碩士專業(yè)：軟件工程指導(dǎo)教師：吳國新；董永強(qiáng)；樂寧201110東南大學(xué)碩：1：學(xué)位論文AbstractThemainDurDoseofthisthesisistoextractthec010r—markedareasinadocumentimageBasedonthestudyofthedigitalimageprocessingtechnology

2、，themethodsofextractingthecolor—markedareasarepresentedinthisthesisInthisthesis，thegeneralproceduresofthedigitalimageprocessingandtheclassicalalgorithIIIsofthedocumentimageprocessingarediscussedfirstlyThen。theproblemofex

3、tractingthecolor—markedareasisdividedintothreeStepS：pre—pr0CeSSingthedOCUment1mage，extraCtlngandrepreSentlngC0上Orareas，calculatingfeaturesandidentifyingtheareas1Pre—processingthedocumentimage：Thissectionfocusesonthedocum

4、entimagebinarizationBasedontheclassicalalgorithmsofthebinarization，animprovedapproachisproposed，inwhichtheoriginalimageissegmentedintothecolorareasandthenon—coloredareasfirstlyThen，twoglobalthresholdsareselectedfortheare

5、asbinarization2Extractingandrepresentingcolorareas：Thissectionfocusesonextractingthecolorareasandrepresentingtheareasusingrectangularp01ygonsTwomethodsareproposedforextractingthecolorareasbasedonregionlabelingandedgedete

6、ctionrespectivelyThefirstmethoddividesthecolorareasintobandregionsThenitextractsthebandregions，andmergesthebandregionsintectangularpolygonsThesecondmethoddetectstheedgesofthecolorareasusingthe1inedetectionalgorithmfromRa

7、faelThenitcombinestheedgesintobandregions，andmergesthebandregionsintectangularp01ygons3Calculatingfeaturesandidentifyingtheareas：Thissectionfocusesontheproblemofdistinguishingthecolor—markedareasandthecolorimageonesByana

8、lyzingthefeaturesofdifferentkindsofareas，twomethodsofidentifyingtheareasareproposedThefirstmethodcalculatesthecolorfeaturesinHSVandRGB，andthenidentifiestheareasbythevarianceofthecolorfeaturesThesecondmethodcalculatesthed

9、ensityofthecolormark’spixelsandthetext’spixels，andthenidentifiestheareascombinedwiththeGeometricinformationoftheareasTovalidatetheproposedmethods，alargenumberofdocumentimagesselectedfromtheimagedatabaseareprocessedbythep

10、roposedmethodsandtheScaleSaliencyalgorithmTheexperimentalresultsshowthatthemethodsproposedbythisthesisarebetterthantheScaleSaliencyalgorithmFurthertestsshowthattheproposedmethodscanbeappliedtomostofthedocumentimagesTheme

11、thodbasedonregionlabelingismoresuitableforthedocumentimagescontainingthedotted—line—markedareasThemethodbasedonedgedetectionismoresuitableforthedocumentimagecontainingcolorimageareasKeywords：Documentimage，Areaextraction，