Seite - 203 - in Document Image Processing

J. Imaging 2018,4, 32 of samplesprovidedto theCAE,which leads tomorerobust featurerepresentationsandsubsequently betterclassificationresults. 6.Conclusions In thispaper,wehavepresentedanewversionof theAcTiVdataset for thedevelopmentand evaluationof textdetectionandrecognitionsystemstargetingArabicnewsvideo. Thisdataset is freely available toresearch institutions.Wehaveprovideddetailsabout thecharacteristicsandstatisticsof thedatabase.Wehavealsoreportedaboutourground-truthingsoftwareusedtosemi-automatically annotate thevideo clips andouropen textdetection evaluation tool. Wehave evaluatedfive text detectionandrecognitionalgorithmsasproof-of-conceptof thenewdataset. Additionally, a setof evaluationprotocolshasbeenmadetomeasure thesystemsperformanceunderdifferentsituations. Theexperimental resultshaveshownthat there isstill roomfor improvement inbothdetectionand recognitionofArabicvideo text.Welookforwardtomoreresearchers joining thechallengingresearch topicofArabicvideotextsdetectionandrecognition. Acknowledgments: The researcherswould like to thankall theTVchannels for providinguswithdata and multimediafiles, especially thearchivestaffofElWataniya1 (TunisiaNat1)TV. AuthorContributions:OussamaZayeneconceivedthedataset,developedthe toolsandrealizedtheexperiments under the supervision andhelp of professorsNajouaEssoukri BenAmara, JeanHennebert andRolf Ingold. SamehMasmoudiTouj contributedto thecollect,designandannotationof thedataset, andverifiedtheannotated data.All theco-authorshavesubstantiallyrevisedthemanuscript. Conflictsof Interest:Theauthorsdeclarenoconflictof interest. References 1. Lu,T.;Palaiahnakote,S.;Tan,C.L.;Liu,W.VideoTextDetection; SpringerPublishingCompany, Incorporated: London,UK,2014. 2. Ye, Q.; Doermann,D. Text detection and recognition in imagery: A survey. IEEETrans. PatternAnal. Mach. Intell. 2015,37, 1480–1500. 3. Yin,X.C.;Zuo,Z.Y.;Tian,S.;Liu,C.L. TextDetection,TrackingandRecognition inVideo:AComprehensive Survey. IEEETrans. ImageProcess. 2016,25, 2752–2773. 4. Lienhart,R. VideoOCR:Asurveyandpractitioner’sguide. InVideoMining; Springer: Boston,MA,USA, 2003;pp. 155–183. 5. Yang, H.; Quehl, B.; Sack, H. A framework for improved video text detection and recognition. Multimed.ToolsAppl. 2014,69, 217–245. 6. Poignant, J.;Bredin,H.;Le,V.B.;Besacier,L.;Barras,C.;Quénot,G. Unsupervisedspeaker identification usingoverlaidtexts inTVbroadcast. InProceedingsoftheInterspeech2012—ConferenceoftheInternational SpeechCommunicationAssociation,Portland,OR,USA,9–13September2012;p.4. 7. Märgner,V.;ElAbed,H.Guide toOCRforArabicScripts; Springer: Berlin,Germany,2012. 8. Touj, S.M.;Amara,N.E.B.;Amiri,H. ArabicHandwrittenWordsRecognitionBasedonaPlanarHidden MarkovModel. Int.Arab J. Inf. Technol. 2005,2, 318–325. 9. Lorigo,L.M.;Govindaraju,V. OfflineArabichandwritingrecognition:Asurvey. IEEETrans. PatternAnal. Mach. Intell. 2006,28, 712–724. 10. Chammas,E.;Mokbel,C.;Likforman-Sulem,L.Arabichandwrittendocumentpreprocessingandrecognition. InProceedingsof the201513thInternationalConferenceonDocumentAnalysisandRecognition(ICDAR), Tunis,Tunisia,23–26August2015;pp. 451–455. 11. Jamil,A.; Siddiqi, I.;Arif,F.;Raza,A. Edge-basedfeatures for localizationofartificialUrdutext invideo images. InProceedingsof the2011InternationalConferenceonDocumentAnalysisandRecognition,Beijing, China,18–21September2011;pp. 1120–1124. 12. Halima, M.B.; Karray, H.; Alimi, A.M. Arabic text recognition in video sequences. arXiv 2013, arXiv:preprint/1308.3243 203