基于非線性方程的圖像處理算法設(shè)計.pdf

1、南京理工大學(xué)碩士學(xué)位論文基于非線性方程的圖像處理算法設(shè)計姓名：陳宏君申請學(xué)位級別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：王執(zhí)銓茅耀斌20050615南隸理工文學(xué)碩士拳位論文基于菲線性方程豹蕊像懿聯(lián)舞法最計AbstractThepresentdissertationhasappbednonfinearequationsfromtheviewpointofscale—spaceintheresearchofimagefiltering，se

2、gmentationandvideotrackingThemaincontributionsofthedissertationareitemizedasfollows：(1)Animprovedimagefilteringalgorithmbasedonfourthorderpartialdifferentialequation(PDE)isputforthThealgorithmcouldsimultaneouslysuppressm

3、ixingGaussianandsaltandpeppertypenoiseswhileeliminatetheblockingeffectderivedfromloworderPDEs；(2)Togetridofnoiseswithlargevariance，anonlinearadaptivediffusioniterativefilterisimposedonhighfrequencycoefficientsofimagesaft

4、erwaveletdecomposition，whichhasachievedsatisfactoryresults；(3)Animageshockfilteringalgorithmemployingnonlinearcomplexequationsisproposed。whichcarteffectivelysmoothnoiseswhile豫ha蘸eeimageedges；(4)Furthermore，theCGLequation

5、thatusedtodescribethesuperconductorinitscriticalphasehasbeenincorporatedintocomplexdiffusionequationtOformashockfilteringmodelthathastheabilityofenhancingedgesintheprocessofnoiseelimination；(5)Acolorimagefilteringalgorit

6、hrabasedonBeltramiFlowhasbeenimporved。Byusingentropyconservationdiseretizationmethod，thealgorithmhasanisotropicdiffusingspeedingradientdirectionandcontourdirectionSOthatmixingnoisewithGaussianandSaltandPeppertypecouldbee

7、fficientlysuppressedinonlyafewroundsofiterations；(◇Mterimagesegmentationaigorithmsbasedonspatialscale—spaceissummarized，onefastedgedetectionmethodusingcomplexdiffusionequationhasbeenprovidedwhoselocalizationprecisionando

8、perationspeedarebothimproved；(7)OnefeasiblesegmentationalgorithmwithgoodsegmentationresultbutlowoperationspeedisalsobroulghtoutinthethesisThelevelsetmethodinthecomplexnonlinearscale—spacehasbeenusedinthealgorithm；(8)Onea

9、ctivecontourvideotrackingmethodblendingchaos，geneticalgorithmandgreedyalgorithmisproposed，whosesegmentationresulthasexceededthatofthemethodsjustUSeSgreedyalgorithmintermsofsingleframesegmentationFinallyafteranalyzingthed