Presentation is loading. Please wait.

Presentation is loading. Please wait.

WP3: Image Segmentation - OCR Stavros Perantonis, Vassilis Maragos Edinburgh, March 6-7, 2003 Institute of Informatics & Telecommunications NCSR “Demokritos”

Similar presentations


Presentation on theme: "WP3: Image Segmentation - OCR Stavros Perantonis, Vassilis Maragos Edinburgh, March 6-7, 2003 Institute of Informatics & Telecommunications NCSR “Demokritos”"— Presentation transcript:

1 WP3: Image Segmentation - OCR Stavros Perantonis, Vassilis Maragos Edinburgh, March 6-7, 2003 Institute of Informatics & Telecommunications NCSR “Demokritos”

2 © NCSR, Edinburgh, March 6-7, 2003 Banner Recognition

3 © NCSR, Edinburgh, March 6-7, 2003 Banner characteristics -Low resolution -Graphics, noiseless -Anti-aliasing -Color contrast visible by the human eye -Text body of restricted thickness

4 © NCSR, Edinburgh, March 6-7, 2003 OCR Input Original image: B/W OCR input: B/W OCR input after Text Area Enhancement Pre-processing Tool:

5 © NCSR, Edinburgh, March 6-7, 2003 Text Area Enhancement Pre-processing Tool

6 © NCSR, Edinburgh, March 6-7, 2003 Text Area Enhancement Pre-processing Tool

7 © NCSR, Edinburgh, March 6-7, 2003 FineReader5ReadIris7 Text Area Enhancement + FineReader5 Text Area Enhancement + ReadIris7 x1W",hmo".,. ~..~.. u.:.-.é~W."hm.- x2 Watch movies. ~'~..'"...,., Watch movies... x4 Watch movies.. é... x8 Watch movies.. é...

8 © NCSR, Edinburgh, March 6-7, 2003 FineReader5ReadIris7 Text Area Enhancement + FineReader5 Text Area Enhancement + ReadIris7 x1.,....",..II'I;a Novità e offerte x2~(i~ Nevità e Gfferte x4~(3~ Novità € offerte Nevità e efferte x8-- Novità € offerte Nevità e efferte

9 © NCSR, Edinburgh, March 6-7, 2003 FineReader5ReadIris7 Text Area Enhancement + FineReader5 Text Area Enhancement + ReadIris7 x114" 64M' P.Cil33 "1411. SDRAM i x264M:14» 64M' RB113S "1488. SDR4M' ;,;, i OCLIC. "E"E x4 64M: SH3I 14" 'A CLICK MERE e4M. R&138 4~1411 U. IORolM. I ~ I:LII:K HERE x8 Jd CLICK MERE RAM' PC133 D4IVI* sdram CLICK MERE. RGJ133 tt~. SOR.IM Il I )111 E:LIE:K HEFi!E Result3

10 © NCSR, Edinburgh, March 6-7, 2003 Next Steps  Automatic Evaluation of the Text Area Enhancement Pre- processing Tool (create Ground Truth Annotations – record improved Recognition Rates for letters/words).  Parameters fine tuning for the Text Area Enhancement Pre- processing Tool (resolution, number of iterations)  Select the appropriate OCR engine.  Train the OCR engine for better results.  Add CROSSMARC lexicons - Post-processing technique to increase recognition accuracy  Integration with NERC, Delivery of an Ellogon-based application

11  Unsurpassed Accuracy. Thanks to its use of IPA Technology, FineReader has an unprecedented recognition accuracy. FineReader has come out on top in comparative tests.  Impeccable Layout Retention. New recognition procedures retain the look and feel of your printed documents, be it wrap-around text, vertical text, columns, tables, non-rectangular pictures or varying fonts. Wide range of document saving formats is supported.  PDF Input and Output. Recognize, edit and save documents in PDF format. Dozens of multilanguage fonts included!  Full HTML Support.  FineReader is a Pleasure to Use.  Batch Document Support provides you with the tools you need to work with multipage documents.  The Spelling-check system  Multilingual Document Recognition. FineReader is the leading multi-national OCR software. It recognizes texts in 122 languages  Quick Export to Microsoft Word, Excel and Outlook. FineReader 6.0: Key Features www.finereader.com

12  Unmatched Combination of Accuracy and Speed. Less editing, increase in performance.  PDF Input. Open PDF documents (even read-only!), and convert them into editable files you can send directly to your favorite application.  Page Orientation and Image Deskew. Automatically detects the document orientation and the text skew.  Powerful Adjust Image Option. Restore degraded documents with manual or automatic image adjustments and despeckling options.  Color Document Recognition. Recognizes color documents and text on colored backgrounds. Retains any pictures in color on the output file.  Foreign Language Support. Recognizes up to 104 different languages:  New User Interface. The new user-friendly interface includes a redesigned thumbnail bar and guides you intuitively through the different recognition steps.  Flowing Text Mode. Thanks to the powerful Autoformat™ technology, pictures, graphics and tables are positioned correctly and the text nicely flows accross columns or pages.  New "Send To" Mode. The new “Send To” mode automatically sends the output result to the selected application such as Microsoft® Word, Microsoft® Excel, etc.  Multipage documents/batch OCR ReadIris 8: Key Features www.irislink.com


Download ppt "WP3: Image Segmentation - OCR Stavros Perantonis, Vassilis Maragos Edinburgh, March 6-7, 2003 Institute of Informatics & Telecommunications NCSR “Demokritos”"

Similar presentations


Ads by Google