摘要
Anewapproachbasedonmultiwaveletstransformationandsingularvaluedecomposition(SVD)isproposedfortheclassificationofimagetextures.LowersingularvaluesaretruncatedbasedonitsenergydistributiontoclassifythetexturesinthepresenceofadditivewhiteGaussiannoise(AWGN).Theproposedapproachextractsfeaturessuchasenergy,entropy,localhomogeneityandmax-minratiofromtheselectedsingularvaluesofmultiwaveletstransformationcoefficientsofimagetextures.Theclassificationwascarriedoutusingprobabilisticneuralnetwork(PNN).Performanceoftheproposedapproachwascomparedwithconventionalwaveletdomaingraylevelco-occurrencematrix(GLCM)basedfeatures,discretemultiwaveletstransformationenergybasedapproach,andHMMbasedapproach.Experimentalresultsshowedthesuperiorityoftheproposedalgorithmswhencomparedwithexistingalgorithms.
出版日期
2007年04月14日(中国期刊网平台首次上网日期,不代表论文的发表时间)