基于模糊马尔科夫随机场的遥感图像分割算法研究

《基于模糊马尔科夫随机场的遥感图像分割算法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、ABSTRACTABSTRACTRemote-sensingimagesegmentationistheprocesstodividetheimageintoregions withdifferentfeaturesandextracttheobjectivethroughsegmentation，whichwill probablybeusedintothenextprocessingstep.Remote-sensingimagesegmentationis theimportantstepfromima

2、geprocessingtoimageanalysis.Inrecentyears,theunsupervisedsegmentationalgorithmsconstrainedbyMRFas theprioriprobabilitymodelarewidelyusedintheimagesegmentation.The experimentsshowthatthemodelcansignificantlyimprovethequalityoftheimage segmentation.Asaresulto

3、finfluencesoftheenvironmentandsensors,thecharactersof remote-sensingimagesarehighgraylevel,sufficienttextureandfaintborder，apply standardMRFtothespecklenoisesinsegmentationcan’tgivesatisfyingresults.Aimed atthehighlevelsofspecklenoisesandfaintborders,inthis

4、paper,weapplythenew unsupervisedalgorithmbasedonFuzzyMarkovrandomfieldtoimagesegmentation. TheFuzzyMarkovrandomfieldfusesthefuzzyset-basedandtheMarkovmodel-based methods.ThesegmentationexperimentsofSARimagesdemonstratethattheproposed algorithmiscorrect.Them

5、ainworkandinnovationsofthispaperisunderfollowing: (1)Inthisarticle,twomainkindsofthealgorithms,fuzzyset-basedandMarkov model-basedalgorithms,arediscussed.Wecontrasedthetwoalgorithmsin remote-sensingimagesegmentation.(2)Inordertosolverespectiveinsufficiencyo

6、fabovetwomethods,theFuzzy Markovrandomfieldisappliedtoremote-sensingimagesegmentation.Frameworkin segmentationalgorithmbasedonFuzzyMarkovrandomfieldisproposed.Everypixel inimageisnotassignedacertainclassbutafuzzyclass.Basedonthefuzzyrandom variables,themode

7、lcombinestherandomnessandfuzzinessinimagesegmentation.In termsofasoftsegmentationalgorithm,themorereasonablytoreceivethepriori probabilityinaccordancewiththefeaturesoftheimage,themorecorrectthepriori probabilityistobeintheimagesegmentation.(3)Whenwedeviseth

8、ecliquepotentialfunctioninFMRF,wedevelopedthe FuzzyMLLmodelbasedontraditionalMLL.ThecliquepotentialfunctioninFuzzyIIIABSTRACTmodelstartsfromthedistanceoftwopointsandtheclassicalcliquepotentialfunctionb