分水岭算法MATLABWatershedalgorithmMATLAB文档格式.docx

分水岭算法MATLABWatershedalgorithmMATLAB文档格式.docx

《分水岭算法MATLABWatershedalgorithmMATLAB文档格式.docx》由会员分享，可在线阅读，更多相关《分水岭算法MATLABWatershedalgorithmMATLAB文档格式.docx（13页珍藏版）》请在冰豆网上搜索。

%imcloseandimreconstruct,imcomplement,imregionalmax,bwareaopen,graythresh,imimposeminfunctionandsoon.

ReadintheColorImageandConvertittoGrayscale

%thefirststep:

readthecolorimageandconvertittograyscaleimage

CLC.Clearall;

Closeall;

RGB=imread（'

pears.PNG'

）;

Ifndims（RGB）==3

I=rgb2gray（RGB）;

Theelse

I=RGB.

Theend

Figure（'

units'

'

position'

[0,0,1]）;

Subplot（1,2,1）;

Imshow（RGB）;

Thetitle（'

artwork'

Subplot（1,2,2）;

Imshow（I）;

Title（'

grayscale'

Step2:

%UsetheGradientMagnitudeastheSegmentationFunction

%2:

thegradientisusedasapartitionfunction

%UsetheSobeledge,imfilter,andsomesimpleproductstocomputethegradient.

ThegradientishighatThebordersofTheobjectsandlow（mostly）insideTheobjects.

%usetheSobeledgeoperatorforthehorizontalandverticaldirectionfilter,andthencalculatethemodulusvalue,Sobeloperatortofiltertheimageintheborderwillbedisplayedafterthelargervalue,intheabsenceofbordervaluewillbesmall.

Hy=fspecial（'

sobel'

Hx=hy'

;

Iy=imfilter（I）,hy,'

get'

Ix=imfilter（I）,hx,'

ream'

Gradmag=SQRT（Ix.^2+Iy.

^2）;

Imshow（I,[]）,title（'

grayscaleimage'

）

"

Imshow"

（gradmag,[]）,title（'

gradientimage'

Canyousegmenttheimagebyusingthewatershedtransformdirectlyonthegradient?

CanIusewatershedalgorithmdirectlyforgradientimageimage?

L=watershed（gradmag）;

Lrgb=label2rgb（L）;

Imshow（gradmag,[]）,title（'

Imshow（Lrgb）;

gradientamplitudeforwatershedtransformation'

%No.Withoutadditionalpreprocessingtotheasthemarkercomputationsbelow,usingthewatershedtransformdirectly,theresultsin"

oversegmentation."

Theresultsofthewatershedalgorithmcanbeover-segmentedbyusingtheimageofgradientmodulus.Therefore,itisusuallynecessarytomarktheforegroundobjectsandbackgroundobjectsseparatelytogetabettersegmentationeffect.

Step3:

%MarktheForegroundObjects

marktheforegroundobject

%Avarietyofprocedurescouldbeappliedheretofindtheforegroundmarkers,

%whichmustbeconnectedblobsofpixelsinsideeachoftheforegroundobjects.

Thisexampleyou'

llusemorphologicaltechniquescalled"

opening-byreconstruction"

and"

closing-by-

reconstruction"

to"

clean"

uptheimage.

Theseoperationswillcreateflatmaximainsideeachobjectthatcanbelocatedusingimregionalmax.

Therearemultiplewaystoapplyheretoobtaintheforegroundmarkup,whichmustbetheconnectiondotpixelswithintheforegroundobject.Inthisexample,wewillusemorphologicaltechniques"

basedonopenreconstruction"

closedreconstruction"

tocleanupimages.

Theseoperationswillcreateunitsofmaximumvaluewithineachobject,enablingtheuseofimregionalmaxtolocate.

%openoperationandclosedoperation:

aftercorrosion,theexpansioniscalledopen;

Afterexpansion,corrosioniscalledclosure.Openandclosedoperationscanremovespecificimagedetailsthataresmallerthanstructuralelements,whileensuringthatthereisnoglobalgeometricdistortion.

The%openoperationcanfilteroutthelittlespikesthataresmallerthanthestructuralelements,cuttingoffthelong,thinonesandseparatingthem.Theclosedoperationcanconnectthesmallgaporholefillofthestructureelement,andconnecttheshortintervals.

Itisanerosionfollowedbyamorphologicalreconstruction.

Let'

scomparethetwo.

First,computetheopeningusingimopen.

%openoperationisanexpansionaftercorrosion,basedonopenreconstruction（basedontheoperationofreconstruction）,aftercorrosion,morphologicalreconstructioniscarriedout.Let'

scomparethesetwoways.First,openanoperationwithimopen.

Se=strel（disk,20）.

Io=imopen（I,se）;

Imshow（I,[]）;

Imshow（Io）,title（'

imageoperation'

%Nextcomputetheopening-by-reconstructionusingimerodeandimreconstruct.

%next,buildontheopenreconstructioncalculationbycorrosion.

Ie=imerode（I,se）;

Iobr=imreconstruct（Ie,I）;

Imshow（Iobr,[]）,title（'

basedonopenreconstructedimage'

Followingtheopeningwithaclosingcanremovethedarkspotsandstemmarks.Comparearegularmorphologicalclosingwithaclosing-by-remission.

Afterthe%openoperation,theclosedoperationcanberemovedtoremovethedarkerspotsandboughs.Contrastthenormalmorphologyclosureandtheclosed-basedreconstructionoperation.First,useimclose:

TheIoc=imclose（Io,se）;

Ic=imclose（I,se）;

Subplot（2,2,1）;

Subplot（2,2,2）;

Imshow（Io,[]）;

openoperationimage'

Subplot（2,2,3）;

Imshow（Ic,[]）;

closedoperationimage'

Subplot（2,2,4）;

Imshow（Ioc,[]）,title（'

openclosedoperation'

%Nowuseimdilatefollowedbyimreconstruct.Noticeyoumustcomplementtheimageinputsandtheoutputofimreconstruct.

ThecomplementoftheimageIMcanbeabinary,intensity,orRGBimage.im2hasthesameclassandsizeasIM.

%nowuseimdilate,thenusetheimreconstruct.Attentionmustbecomplementarytotheinputimages,forcomplementaryimreconstructoutputimages.

%IM2=imcomplement（IM）tocomputethecomplementofimageIM.IMcanbebinaryimage,orRGBimage.IM2andIMhavethesamedatatypeandsize.

Iobrd=imdilate（Iobr,se）;

Iobrcbr=imreconstruct（imcomplement（Iobrd）,imcomplement（Iobr））;

Iobrcbr=imcomplement（Iobrcbr）;

normalized'

theposition'

[0011））;

Imshow（Ioc,[]）;

openandclosedoperation'

Imshow（Iobr,[]）;

Imshow（Iobrcbr,[]）,title（'

basedonclosedreconstructionimages'

YoucanseebycomparingIobrcbrwithIoc,reconstructionandclosingaremoreeffectivethan

Attheendoftheday,thewholeoftheobjectswasmadeupofthewholeoftheobjects.

%CalculatetheregionalmaximaofIobrcbrtoobtaingoodforegroundmarkers.

BycomparingIobrcbrandloc,itcanbeseenthatunderthe

applicationofremovingthesmudgewithoutaffectingtheglobalshapeoftheobject,theopenandclosedoperationbasedonreconstructionismoreeffectivethanthestandardopenandclosedreconstruction.CalculatethelocalmaximsofIobrcbrtogetabetteroutlook.

FGM=imregionalmax（Iobrcbr）;

Subplot（1,3,1）;

Subplot（1,3,2）;

Imshow（Iobrcbr,[]）;

openclosedoperationbasedonreconstruction'

Subplot（1,3,3）;

Imshow（FGM,[]）;

localmaximumimage'

%Tohelpinterprettheresult,superimpose（stack）theforegroundmarkerimageonthetheoriginalimage.

%tohelpunderstandthisresult,thestackforegroundismarkedtotheoriginal.

It1=RGB（:

:

1）;

It2=RGB（:

2）;

It3=RGB（:

3）;

It1（FGM）=255;

It2（FGM）=0;

It3（FGM）=0;

I2=cat（3,It1,It2,It3）;

Imshow（RGB,[]）;

originalimage'

Imshow（I2）;

localmaximumsuperimposedtotheoriginalimage'

%Noticethatsomeofthemostly-occludedandShadowedobjectsarenotmarked,whichmeansthattheseobjectswillnotbesegmentedproperlyintheendresult.

%oftheforegroundmarkersinsomeobjectsgorightuptotheobjects'

edge.Thatmeansyoushouldcleantheedgesofthemarkerblobsandthenshrinkthemabit.

Youcandothisbyaclosingfollowedbyanerosion.

%notethatmostclos