AlgorithmDescriptionRecognizingobjectsfromlargeimagedatabases,histogrambasedmethodshaveprovedsimplicityandusefulnessinlastdecade.Initially,thisideawasbasedoncolorhistogramsthatwerelaunchedbyswain[1].Thisalgorithmpresentsthefirstpartofourproposedtechniquenamedas“HistogramprocessedFaceRecognition”[2]Fortraining,grayscaleimageswith256graylevelsareused.Firstly,frequencyofeverygray-leveliscomputedandstoredinvectorsforfurtherprocessing.Secondly,meanofconsecutiveninefrequenciesfromthestoredvectorsiscalculatedandarestoredinanothervectorsforlateruseintestingphase.Thismeanvectorisusedforcalculatingtheabsolutedifferencesamongthemeanoftrainedimagesandthetestimage.Finallytheminimumdifferencefoundidentifiesthematchedclasswithtestimage.Recognitionaccuracyisof99.75%(onlyonemis-matchi.e.recognitionfailsonimagenumber4ofsubject17)[1]M.J.SwainandD.H.Ballard,“Indexingviacolorhistogram”,InProceedingsofthirdinternationalconferenceonComputerVision(ICCV),pages390–393,Osaka,Japan,1990.[2]Fazl-e-Basit,YounusJavedandUsmanQayyum,"FaceRecognitionusingprocessedhistogramandphaseonlycorrelation",3rdIEEEInternationalConferenceonEmergingTechnologypp.238-242
1