Seite - 43 - in Document Image Processing

J. Imaging 2018,4, 80 andtheBH2MDBandtheWashingtonDBtoprovetheapplication’sutilityonhistoricaldocuments. TheresultsprovedquitesatisfactorywithaRMSEof less thanthree for thePrintDB, less thanfour for theWashingtonDB, less thanfiveforBH2M,andless thaneight for thenormalslanteddocumentsof theTrigraphSlant. The improvement inourword-spottingsystemfordifficulthistoricaldocuments was impressive. The technique fails if thecharactermainbodysizedetection isnot correct. Thus, agoodmain bodysizedetectionalgorithmisrequired.Moreover, theproposedtechnique isappropriateonly if the slant ishomogenous throughout theentiredocument image.However,moreslant removalalgorithms couldbeusedincombinationwith theproposedtechnique. This isamongour futureplans. AuthorContributions:Thementionedauthors,E.K.,L.L.-S.andN.V.havecontributedtoall stagesof thework. Funding:This researchreceivednoexternal funding. Conflictsof Interest:Theauthorsdeclarenoconflictof interest. References 1. Parvez,T.M.;Sabri,A.M.Arabichandwritingrecognitionusingstructuralandsyntacticpatternattributes. PatternRecognit. 2013,46, 141–154. [CrossRef] 2. José, A.R.-S.; Perronnin, F. Handwritten word-spotting using hidden Markov models and universal vocabularies.PatternRecognit. 2009,42, 2106–2116. 3. Brink,A.A.;Niels,R.M.J.; vanBatenburg,R.A.;vandenHeuvel,C.E.; Schomaker,L.R.B.Towardsrobust writerverificationbycorrectingunnatural slant.PatternRecognit. Lett. 2011,32, 449–457. [CrossRef] 4. Bozinovic,R.; Srihari, S.Off-linecursivescriptwordrecognition. IEEETrans. PatternAnal.Mach. Intell. 1989, 11, 68–83. [CrossRef] 5. Kim, G.; Govindaraju, V. A lexicon driven approach to handwritten word recognition for real-time applications. IEEETrans. PatternAnal.Mach. Intell. 1997,19, 366–379. 6. Shridar,M.; Kimura, F.Handwritten address interpretation usingword recognitionwith andwithout lexicon. In Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, Vancouver,BC,Canada,22–25October1995;Volume3,pp.2341–2346. 7. Papandreou,A.; Gatos, B.Word slant estimation using non-horizontal character parts and core-region information. InProceedings of the 10th IAPR InternationalWorkshoponDocumentAnalysis Systems (DAS2012),GoldCoast,QLD,Australia, 27–29March2012;pp.307–311. 8. Alessandro, V.; Luettin, J. A new normalization technique for cursive handwritten words. PatternRecognit.Lett. 2001,22, 1043–1050. 9. Kavallieratou,E.;Fakotakis,N.;Kokkinakis,G.SlantestimationalgorithmforOCRsystems.PatternRecognit. 2001,34, 2515–2522. [CrossRef] 10. Britto,A., Jr.; Sabourin,R.;Lethelier,E.;Bortolozzi,F.; Suen,C. Improvementhandwrittennumeral string recognitionby slant normalization and contextual information. InProceedings of the 7th International WorkshoponFrontiers inHandwritingRecognition,Amsterdam,TheNetherlands,11–13September2000; pp.323–332. 11. Ding,Y.;Kimura,F.;Miyake,Y.;Shridhar,M.Accuracy improvementofslantestimationforhandwritten words. In Proceedings of the International Conference on Pattern Recognition, Barcelona, Spain, 3–7September2000;Volume4,pp.527–530. 12. Ding, Y.; Ohyama,W.; Kimura, F.; Shridhar,M. Local slant estimation for handwritten Englishwords. In Proceedings of the 9th InternationalWorkshop on Frontiers inHandwriting Recognition (IWFHR), Kokubunji,Tokyo, Japan,26–29October2004;pp.328–333. 13. Bertolami,R.;Uchida,S.;Zimmermann,M.;Bunke,H.Non-uniformslantcorrectionforhandwritten text linerecognition. InProceedingsof the9th InternationalConferenceonDocumentAnalysisandRecognition, Parana,Brazil, 23–26September2007;pp.18–22. 14. Taira,E.;Uchida,S.; Sakoe,H.Non-uniformslantcorrectionforhandwrittenwordrecognition. IEICETrans. Inf. Syst. 2004,E87-D, 1247–1253. 43