MostexistingapplicationsofcentroidalVoronoitessellations(CVTs)lackconsiderationofthelengthoftheclusterboundaries.Inthispaperweproposeanewmodelandalgorithmstoproducesegmentationswhichwouldminimizethetotalenergy—asumoftheclassicCVTenergyandtheweightedlengthofclusterboundaries.TodistinguishitwiththeclassicCVTs,wecallitanEdge-WeightedCVT(EWCVT).TheconceptofEWCVTisexpectedtobuildamathematicalbaseforallCVTrelateddataclassificationswithrequirementofsmoothnessoftheclusterboundaries.TheEWCVTmethodiseasyinimplementation,fastincomputation,andnaturalforanynumberofclusters.
