Geometric layout analysis of scanned documents [Elektronische Ressource] / by Faisal Shafait

Geometric Layout Analysisof Scanned DocumentsDissertationsubmitted to theDepartment of Computer ScienceTechnical University of Kaiserslauternfor the fulfillment of the requirements for the doctoral degreeDoctor of Engineering(Dr.-Ing.)byFaisal ShafaitThesis supervisors:Prof. Dr. Thomas Breuel, TU KaiserslauternProf. Dr. Horst Bunke, University of BernChair of supervisory committee:Prof. Dr. Karsten Berns, TU KaiserslauternKaiserslautern, April 22, 2008D 386AbstractLayout analysis–the division of page images into text blocks, lines, and determinationof their reading order–is a major performance limiting step in large scale document dig-itization projects. This thesis addresses this problem in several ways: it presents newperformance measures to identify important classes of layout errors, evaluates the per-formance of state-of-the-art layout analysis algorithms, presents a number of methods toreduce the error rate and catastrophic failures occurring during layout analysis, and de-velopsastatisticallymotivated, trainablelayoutanalysissystemthataddressestheneedsof large-scale document analysis applications. An overview of the key contributions ofthis thesis is as follows.First, thisthesispresentsanefficientlocaladaptivethresholdingalgorithmthatyieldsthe same quality of binarization as that of state-of-the-art local binarization methods,but runs in time close to that of global thresholding methods, independent of the localwindow size.