B08 - last change: 17-01-2007
BOBCATSSS 2008
Providing Access to Information for Everyone
| Speakers | |
|---|---|
|
Dimitar Poposki |
| Schedule | |
|---|---|
| Day | 3 |
| Room | Donat Exhibition Area |
| Start time | 16:00 |
| Duration | 01:30 |
| Info | |
| ID | 229 |
| Event type | Poster |
| Track | Poster |
| Language | English |
Impossible & Advanced Optical Character Recognition
Using a Cheap Point & Shoot Digital Camera
This study was conducted using a point & shoot digital camera Konica Minolta DiMAGE X1 with 8.1 Megapixels with no previous adjustment at 3264x2448 pixels with 72 dpi. Samples from 4 books were used:
- [MILTON, JOHN.] The Grand Case of Conscience concerning the Engagement Stated & Resolved. printed in London 1650
- [PETKOV-MISIRKOV, KRSTE.] Za makedonskite raboti., printed in Sofia 1903
- [DE MONTESQUIEU, MONSIEUR.] Oeuvres. Tome I, printed in London 1787
- [VARIOUS AUTHORS] Macedonian Review, year IV, book#4, printed in Sofia 1933 Only JONH MILTON book was “scanned” at Alexander Library at Rutgers, the State University of New Jersey with expensive professional camera scanner. It is used here for the sake of the comparison in relation to the curve of the pages and the high resolution professional cameras Vs cheap point & shoot digital cameras. The "scanning" was performed using a light bulb of 50W in room conditions. Some pattern training was done using Abbyy Finereader 6.0. Different compression methods were used to reduce the file size to its minimum web centricity. Optical Character Recognition engines require the minimum of 150 dpi (the optimum is 300 dpi) for accurate results. Using an experimental program with Abbyy Finereader 5.0 engine at 72 dpi JPEG (instead of TIF) files some amazing results appeared.