2008-IUPR-21Jan_0824.pdf

0.0 0.00 0.000 0.001 0.002 0.005 0.010 0.020 0.050 0.083 0.1 0.100 0.200 0.26 0.34 0.463 0.47 0.5 0.500 0.52 0.561 0.6 0.756 0.8 0.86 0.976 0x000000 0xffffff 1-7 1.0 1.23 1.25 1.3 1.34 1.5 1.64 1.66 1.72 1.73 1.95 10.8 100 100.13 101.61 102 105.99 105443 11.2 11.7 11.94 11.98 1162 1167 1173 1177 118.18 12.2 12.3 12.7 12.9 13.26 13.3 13.4 13.43 13.94 14.2 14.4 14.5 14.57 14.7 14.96 144 15.1 15.2 15.25 15.6 155.88 159 15th 16.29 16.49 1600 165.22 166 17.1 17.4 17.70 172 175 18.27 18.3 18.45 181 182 183 188 188.68 18th 19.32 19.66 19.9 190 197 1982 199 199.89 1992 1993 1994 1995 1996 1997 1998 1999 2-3 2.0 2.345 2.4 2.5 2.578 2.8 2.82 2.9 20.0 200 2000 20000 2001 2002 2003 2004 2005 2006 2007 205 207 21.63 217 22.02 22.07 22.6 224 23.0 23.5 23.54 24-bit 24.4 24.90 242 24247 25000 256 27.43 28.29 28.78 2nd 3.0 3.30 3.4 3.6 3.77 3.92 30.38 300 300-dpi 30000 302 31.09 31.10 311 34.29 34.59 35.22 35000 36.05 362 368 37.15 37.18 37.41 370 379 38.24 38.34 382 39.0 39.05 3rd 4.3 4.4 4.6 4.7 40.18 40.68 400 40000 42.21 42.43 421 424 427 428 43.29 432 433 436 44.13 449 45.73 456 46.2 48.8 5.2 5.3 5.5 5.6 5.8 5.9 500 5010 51.83 52.2 53.03 57.23 5th 6.0 6.2 6.6 6.60 6.7 6.8 6.9 601 604 61.74 621 624 647 651 656 660 67.89 673 68.7 688 689 692 7.5 7.9 74.55 75.6 776 779 799 7th 8.04 8.05 8.3 8.5 8.7 8.9 81.23 819 823 84.16 84.27 872 875 878 88.5 8th 9.1 9.2 9.8 909 91.88 92.32 92.82 93.34 97.17 9703-09 978 98.55 9th ACKNOWLEDGMENTS ALARMS ALGORITHM ALGORITHMS ALL AMD AND ARE ASCII Aachen Aalborg Abstract Adam Additional Additionally After Algorithm Algorithms All Almaden Also Alto Anal Analysis Annapolis Antonacopoulos Applications Applied April Article Artificial Attribute Aug Average BMBF Bachelors Background Baird Bangalore Barcelona Based Before Benchmarking Besides Beusekom Both Bowyer Box Brazil Breuel Bridson Buchman Bunke Bureau CBDAR COLUMN COMPONENTS CONCLUSION CONSIDERED CORRECT CUT Camera-Based Canada Casey Cattoni Census Center Chair Chanda China Chowdhury Cinque Classification Coianiz Communication Computer Conf Conference Consider Constrained Correlation Curitiba Currently Cut D03 D047 DAFS DAS04 DASPerfEv.pdf DEFAULT DETECTION DFKI DFKI-1 DIFFERENT DIRECT DOCUMENTS DRAFT DUMMY Daniel Das Data Default Denmark Department Depending Despite Determining Development Dipl Distance Docstrum Document Doermann Dori Dosch Due Dummy During E-mail EACH ENTRIES ERRORS EURASIP EVALUATION EXPERIMENTS EXPRESSED Each Early Edinburgh Eds Education Eggert Electrical Electronic Empirical Engineering English Equation Evaluation Example FALSE FINDING FOR Faisal False Feb Federal Fig Figure Finally Find First Fisher Fitzgibbon Florence Flynn For Ford Format From Frontiers Furthermore GHz GIVEN GROUND GROUNDTRUTH Gatos Generated Geometric German Germany Gimp GmbH Goldgof Google Gorman Greenbelt Ground Ground-truthing Groundtruth Group Guyon Hamburg Handbook Handwriting Haralick Harvard Hence Heroux High His Hong Hoover However Hull Human IAPR IBM ICDAR IDIAP IEEE II-A II-C IIC III III-A IMAGES INTRODUCTION IPeT IRST IUPR IV-A Image Imaging Index India Information Informative Informtica Instead Institute Instituto Int Intell Intelligence Interestingly International Intl Irniger Italy Iwata Jan Japan Jean-Baptiste Jiang Jose Journal June KEY Kaiserslautern Kanai Kanungo Karatzas Keysers Kise Kong Korea LABELS Labeled Labs Language Layout Layouts Let Levialdi Liang Linux Live Llados Lombardi Lotti MADE MISSED MODIFIED MS-Paint Mach Machine Major Mandal Manhattan Mao Mar March Marinai Marino Marti Massachusetts Mean Messelodi Microsoft Ministry Missed Modena Montreal Multiple NORMALIZED NUMBER Nagy Nartker Nelson New None Nov November Number OCR ONE OPTIMIZED ORIGINAL OVERSEGMENTATIONS OVERSEGMENTED Oct Okun One Only Ontario Opteron Optimal Optimized Original Our Over Oversegmented PAGE PAMI PARAGRAPH PARAMETER PARAMETERS PARC PERCENTAGE PERFORMANCE PHOTOCOPIES PLEASE PSET Page Pakistan Palo Panther Patt Pattern Performance Ph.D PhD Phillips Photocopy Pietikainen Pink Pixel-accurate Politcnica Preparation Previously Princeton Proc Proceedings Processing REFER REFERENCES REPORTED RESULTS RGB RLSA RWTH RXYC Recent Recog Recognition Recommendations Remarkably Rep Representation Representations Research Researchers Retrieval Rice Robust Rosenfeld Ross SAME SCANS SEGMENTATION SEGMENTATIONS SEPARATE SEPARATED SMEARING SPIE Saha San Sanchez Sato Sauvola Scandinavian Science Scientific Secondly Section Segmentation Segmented Seoul Sep Seth Several Shafait Shin Signal Significance Similar Since Singapore Single Six Smearing Soda Software Spain Specification Stdev Such Switzerland Symposium Systems T.M TABLE TERMS TEST TEXT TEXT-LINE TEXT-LINES TEXTLINE-LEVEL THE THREE-COLUMN TOTAL TRAIN TRUTH TUHH TWO TYPES Table Tables Tanimoto Taxila Tech Technical Technology Tecnolgico Terms Test Text-line Textline That The Their Then Theory There Therefore These They This Thoma Thomas Total Total-oversegmentations Train Trans Trento Tsukuba Twenty Two Typically UNDERSEGMENTATIONS UNDERSEGMENTED USA USED UW-I UW-III Undersegmented Understanding University Unlike Unversidad Using Usually VALUES VIII Valencia Valveny Vincent Vision Viswanathan Voronoi Voronoi-based Voronoi-diagram Voronoi-diagram-based Wahl Wang Washington When Whereas Whitespace Wong Word Workshop World X-Y XML Xerox Yanikoglu Yes ZONE ZONE-LEVEL ZONES Zealand Zone Zones a-i abovementioned absolute accepts according accuracy accurately achieve achieved actually adapted add added addition address adds adjacent affected agrees alarm alarms algorithm algorithms allowed allowing allows alternatives altogether alumnus analysis analyze analyzed analyzes analyzing angle appears application application-dependent applicationdependent applications applied applying approach approaches approximation arbitrarily arbitrary architecture archival area areas arise artifacts aspect aspects assess assign assigned associate associated assumption assumptions attention attribute attributes augments author automatic available average avoid axis background badly bar base based baseline basis behavior belong belongs benchmarking benchmarks benefit benefits best better big binary bins bipartite bitmap bitmaps black block block-level blockclassification blocks body book books border borders bottom-up bottomup bound boundaries bounded bounding bounding-box box boxes branch brief broad business calculated called candidate candidates canonical cardinality case cases categorized category certain changed changes channels character characteristics characters check children choice chose chosen circles circumstances citeseer.nj.nec.com classes classification classified classifies clean clearly closest closure clustering coded coincides collection color color-based coloring colors column column-wise columns combinations combined combining come commercial common comparative compare compared comparison competing competition competitionbased competitions complete completely complex component components computation compute computed computing conclude conclusion conducted confidence connected consecutive consider consideration considered consisting consists constrained constraints construct constructed consultant contain contained containing contains content contest continues contributions control convenient correct correctly correlated correlation correspond correspondence correspondences corresponding cost cover covered covers created creating criterion crucial cum cumbersome current cut cuts daniel.keysers data database dataset decades declared decomposed decreases decreasing default define defined defining definitions degradation degree deleted demands demonstrate demonstrates demonstration denote depend dependent depends described describes description desired detailed details detected detection determine determined develop developed deviate deviation dewarping dfki.de diagnosing diagram did differences different difficult digit dimensionless dimensions direction directions directly discuss discussed distance distinct diverse diversity divide divided docstrum document documents does doing dominant dot drawing drawn dummy e.g early ease easily easy edge edges edited effect effective effects embedding emerged empirical enables enclose encoded encoding end ensure entire entirely entities entity environment equal equals error errors especially estimate estimated estimation evaluate evaluated evaluating evaluation evaluations exact exactly example examples exchanged expected expects experiment experimental experimented experiments explain explained extend extent extracted extraction extracts fact factor fail failing fails faisal.shafait fall false far fast features files filled final finding finds flaw flaws focus followed following follows font footer foreground form format formats fraction fractional frame framework free frequently fth ftr fulfil function funded future gain gains gap gave general generally generate generated generating generic geometric given gives giving goal good graph graph-like graphics greater ground ground-truth groundtruth group groups guarantees gutters half handle handwriting happens having head headed header headings heavily height held helps heterogeneous high high-performance higher highlight highlights highly histogram histograms homogeneous honors horizontal horizontally http hypothesis hypothesized i-dots i.e identification identified identifies identify ignored ignoring illustrate illustrating illustration image images imaging implement implementation implemented implements importance important incident include included includes including inconsistencies incorrectly increased increasingly indentation indentations independent index indicated indicates individual individually informatik.uni-kl.de information inliers input inside insights instance instead insufficient intended intention inter-character inter-line inter-word interchange interesting interests interquartile intersected intersection introduce introduced introduces introducing introduction invariant investigated isothetic issue joined journals judging just keeping key kind label labeled labeling labelled labels lack large larger late laude layout layout-based layouts leaf left length letters level levels lie lies like limitation limitations limiting line linearly lines linking literature little locate log2 logical long look lossless low low-resolution lower lowest machine magazines main major majority make makes making manipulation manually margin marginal marks massive masters match matches matching maximal maximum mean meaningful means measure measured measures measuring median mentioned merely merged merging method methodology methods metric metrics mid-point middle million minimum minus mis-leading missed misses missing misssegmentation model modeling models moderately modified multiple named narrow nearest nearest-neighborhood nearly necessarily necessary need needed neighbor neighboring neighbors new node nodes noise non-Manhattan non-overlapping non-text nonManhattan notches noticed notion novel number numbers object object-process objectives observation observations observe observed obstacles obtain obtained occurring off-line offthe-shelf one-column one-to-one operation operator opinion optical optimize optimized optimizing order orientations original outliers output outputs outside over-segment over-segmentation over-segmentations overall overcome overlap overlapping overlaps oversegmentation oversegmentations oversegmented overview page page-segmentation pages pair pairings pairs paper papers paragraph paragraph-level paragraphs parallel parameter parameter-free parameters partial partially participant participate particular particularly parts pass passes past pattern peak percentage perfect perfectly perform performance performed performing performs permits perpendicular photocopied photocopy photocopying physical pictures pixel pixel-accurate pixel-based pixel-correspondence pixel-level pixels plot pointed points polygon polygons popular portability portable pose poses positions possible practice predefined preferable prepared preparing presence present presented presenting presents previous prior probabilistic problem problems process processing producing products professor profile profiles program prohibitive project projection properties property proportion propose proposed prototype proven provides proximity punctuation purpose putting quality quantitative quantitatively quartiles quickly quite range ranking rate rates ratio reading real-world realistic received recognition recognized recommendations recorded rectangle rectangles rectangular recursion recursive recursively reduction refer referred reflect reflected region region-based regions related relationships relative remaining removal removed report reported reports represent representation representations representative represented representing represents requires research researcher resolution resolutions respect respective respectively rest result resulting results retrieval returned returns review right robust root row-wise rule rules run run-length running said sample sampling satisfied satisfies saved scale scaled scan scanned scanning scheme schemes science score scores search sec second section seen segment segmentation segmentations segmented segmenting segments selected senior sensitive sensitivity separate separated separately separating separation separator separators sequence set sets setting severe severity shape shaped shapes shared showed shown shows sides significance significant similar similarities simple simplicity simultaneously single single-column single-digit single-line single-score situation size sizes skew small small-sized smaller smearing smoothed software solution sort sorted sorting spacing spacings spans spatial special specialized specific specified spectrum split splits square stage standard start state-of-the-art statistical step steps stopping stored strength strengths strict strip strong stronger structure structured student studies study style styles subset subsets sufficient suggested suggests suitable summa superfluous support supposed sure survey systems table taken takes tall target task tasks technical techniques tend tends terminals terms test text text-based text-blocks text-column text-columns text-line text-lines text-sized text-zone text-zones textline textline-level theory three-column threshold thresholds tightly time times timing title titles tmb today tolerance took toolkit top-down topological total touches touching tpa tpe training transferring transforms transitive tree tree-based trimming truth tsh tsm tsv ttc tuned turn type types unchanged uncovered under-segmentation under-segmentations under-segmented undersegmentation undersegmentations undersegmented understanding unification unified union unique upper use used useful using usually valid valley valleys value values van vanish variable variation variations varied variety vectorial verified vertical view virtually vision visited visual vol warped way weaknesses weight weighted weighting well-known whiskers white whitespace whitespaces wide widely wider widespread width within-line word words work working works worse www.dsi.unifi.it x-height x-y years yield yielding zero zone zone-level zones