区域生长算法
Abstract
based。nthec。nsistencyrequest。fImagesegmentati。nis
s。mefeaturet。segmentimageint。multipleareas I“sakey
als。abasist。understandingimageanalysis,butinimagestep
itcanpr。Videm。reinf。rmati。nbetter.c。1。rimage,because
thangray
more1evelimage,themoresegmentati。n。f“emhasattracted
methods
fromforandattentions Currently,segmentation
segmentation.
andregiontheareavailablec010rimage1magemostlyexpandedgray—scaleIn“ispaper,thealgorithmbasedonwatershedgrowingisadopted:
tothesmal1regionremoVlnga士ter
algorithmis
regiona1sousedAsremoVingthesmal1regionsgrowing.f0110Ws-
,,
Firstly,watershedalgorithmisusedtoformlngclosed
the
segmentationregions.
anisotropicP—MAsitsoVer—segmentationproblem'isdiffusion
can
soadoptedasforwesmoothlngaspreprocessingwhichedgeinformation,e1iminatenoisetheproblemofllmaIntaln1sandoVer—segmentatlon
effectiVelyimproVed
..,
Secondly,
based
isontheautomaticseedregionselectioninitialsegmentationregions,basedon1sdeslgnedthetheseededre910nandthehlghestselectedmeanhuedifference
similarityishad
Thirdly,fromamongitsadjacentregions regions,there910n
ottheseselectedseedinspectingthe
growingalgorithmisperformedby
itssi叭larltyadjacentregions.Inordertoreducethenumberot
thesmallregionremoVlng
segmentation1ssma上上used,regionsafterregiongrowing,1sandthentheeffectivec010rimageachleVea
alsoisThealgorithmissimulatedbyMATLAB,
nIcomparedwlth