Seite - 13 - in Document Image Processing

J. Imaging 2018,4, 68 taskbecause the imprintsduetoassignedvalues thatarenot inaccordancewith theneighborhood haveunpleasantvisualeffects, anddestroy theoriginal lookof therestoreddocument. Thesimple replacementwiththepredominantgraylevelvalueof thelocalbackgrounddoesnotsolvetheproblem. Toremedythisissue,anon-localgroupbasedadaptivesparseimageinpaintingissuggestedtoestimate plausiblefill-invalues toreplace the identifiedbleed-throughpixels. The inclusionofnon-local similar patches encourages the consistency in localfine textures,withoutblockingor smoothingartefacts. Theproposedimageinpaintingmethodefficientlyemploystheintrinsic localsparsityandthenon-local patchsimilarity. Theperformanceof theproposedmethod is comparedwithother state-of-the-art methodsonadatabaseof recto-versodocumentswithbleed-throughdegradation. AuthorContributions:M.H.conceivedandimplementedthesparserepresentation inpaintingmethod, runthe experimentsandwrote thepaper. A.T. suggested theproblem,devised therestorationalgorithmin itswhole, andcontributedtowrite thepaper. P.S.providedthebleed-throughmapsfor theentiredatasetand,withE.S., improvedthebleed-throughidentificationmethodandoptimizedtherelatedalgorithm.Allauthorsparticipated in theevaluationof theresults, andreadandapprovedthefinalmanuscript. Acknowledgments:ThisworkhasbeenpartiallysupportedbytheEuropeanResearchConsortiumforInformatics andMathematics (ERCIM),within the“AlainBensoussan”FellowshipProgramme. Conflictsof Interest:Theauthorsdeclarenoconflictof interest. References 1. Fadoua,D.;Bourgeois,F.L.;Emptoz,H. RestoringInkBleed-ThroughDegradedDocument ImagesUsing aRecursiveUnsupervisedClassificationTechnique. InDocumentAnalysis SystemsVII; LectureNotes in ComputerScience;Springer: Berlin,Germany,2006;Volume3872,pp. 38–49. 2. Tan,C.L.;Cao,R.; Shen,P. RestorationofArchivalDocumentsUsingaWaveletTechnique. IEEETrans. PatternAnal.Mach. Intell. 2002,24, 1399–1404. 3. Estrada,R.;Tomasi,C.Manuscriptbleed-throughremovalviahysteresis thresholding. InProceedingsof the 10thInternationalConferenceonDocumentAnalysisandRecognition,Barcelona,Spain,26–29July2009; pp.753–757. 4. Shi, Z.; Govindaraju, V. Historical document image enhancement using background light intensity normalization. InProceedingsof the17th InternationalConferenceonPatternRecognition,Cambridge,UK, 26–26August2004;pp.473–476. 5. Tonazzini,A.;Bedini,L.; Salerno,E. Independentcomponentanalysis fordocumentrestoration. Int. J.Doc. Anal. Recognit. 2004,7, 17–27. [CrossRef] 6. Wolf, C. Document ink bleed-through removalwith two hiddenMarkov randomfields and a single observationfield. IEEETrans. PatternAnal.Mach. Intell. 2010,32, 431–447. [CrossRef] [PubMed] 7. Sun,B.;Li,S.;Zhang,X.P.;Sun, J. BlindBleed-ThroughRemoval forScannedHistoricalDocument Image withConditionalRandomFields. IEEETrans. ImageProcess. 2016,25, 5702–5712. [CrossRef] [PubMed] 8. Tonazzini,A. Colorspace transformations foranalysisandenhancementofancientdegradedmanuscripts. J.PatternRecognit. ImageAnal. 2010,20, 404–417. [CrossRef] 9. Drira,F.;Bourgeois,F.L.;Emptoz,H.Restoring InkBleed-ThroughDegradedDocument ImagesUsingaRecursive UnsupervisedClassificationTechnique;Bunke,H.,Spitz,A.,Eds.;Springer: Berlin,Germany,2006;Volume3872, pp.38–49. 10. Tonazzini,A.; Gerace, I.;Martinelli, F. Multichannel blind separation anddeconvolutionof images for documentanalysis. IEEETrans. ImageProcess. 2010,19, 912–925. [CrossRef] [PubMed] 11. Tonazzini, A.; Bedini, L.; Salerno, E. AMarkovmodel for blind image separationby amean-fieldEM algorithm. IEEETrans. ImageProcess. 2006,15, 473–482. [CrossRef] [PubMed] 12. Moghaddam, R.F.; Cheriet,M. Lowquality document imagemodeling and enhancement. Int. J. Doc. Anal.Recognit.2009,11, 183–201. [CrossRef] 13. Moghaddam,R.F.;Cheriet,M. Avariationalapproachtodegradeddocumentenhancement. IEEETrans. PatternAnal.Mach. Intell. 2010,38, 1347–1361. [CrossRef] [PubMed] 14. Tonazzini,A.;Salerno,E.;Bedini,L. Fast correctionofbleed-throughdistortion ingrayscaledocumentsby ablindsourceseparationtechnique. Int. J.Doc.Anal. Recognit. 2007,10, 17–27. [CrossRef] 13