简介:Torealizecontent-basedretrievaloflargeimagedatabases,itisrequiredtodevelopanefficientindexandretrievalscheme.ThispaperproposesanindexalgorithmofclusteringcalledCMA,whichsupportsfastretrievaloflargeimagedatabases.CMAtakesadvantagesofk-meansandself-adaptivealgorithms.Itissimpleandworkswithoutanyuserinteractions.Therearetwomainstagesinthisalgorithm.Inthefirststage,itclassifiesimagesinadatabaseintoseveralclusters,andautomaticallygetsthenecessaryparametersforthenextstage-k-meansiteration.TheCMAalgorithmistestedonalargedatabaseofmorethantenthousandimagesandcompareitwithk-meansalgorithm.Experimentalresultsshowthatthisalgorithmiseffectiveinbothprecisionandretrievaltime.