The following is a superficial analysis of a few OCR packages commercially available. The main objective is to get a sense of the functionality and workflow in existing offerings.
Free / Open Source OCR projects:
- Ocrad (http://www.gnu.org/software/ocrad/)
- jOCR / GOCR (http://jocr.sourceforge.net/)
Commercial OCR software:
- OmniPage v16 - No demo.
- Abbyy FineReader v9 - Windows Only. Trial version
- OcrIRIS Readiris v11 - Windows and MacOS Trial version
OmniPage v16 Pro
Evalutator: Jonathan Hung
Notes:
- Import from scanner, file, or other device.
- Output to a variety of formats: PDF, Searchable PDF, Word, Excel, plain text. But no HTML.
- Software automatically detects text, graphics, and tables on page.
- User can change this by drawing their own bounding boxes.
- Workflow: import, make modifications to layout / regions, convert to text, output to file
- Under more advance "Flexible View", there are tools to rotate and dewarp an image.
- Software attempts to automatically deskew, but you can drag the deskew "mesh" to customize.
- There is a workflow customizer that allows you to define a structured and repeatable process
- Can't apply modifications (i.e. deskewing or cropping) to a single image across multiple images (i.e. no batch process functionality)
- During conversion process, can specify one or multiple columns, presence of tables, forms, etc.
The "Express" interface. Stripped down UI for basic functionality.
The 3D deskew interface under the "Flexible" UI.
Automatic layout detection creates regions and does an error check on converted text.