If the standard OCR SDKs seem not be suited, please provide answers to the listed questions below.
What types of images be processed? (format, resolution)
Where do the images come from?
What kind of information/text should be OCRed?
What Export formats are required? (text only,
XML with coordinates, searchable
PDF, other?)
What is the business model of the solution/device that should use embedded OCR?
What are the quantities the solution/devices should be sold?
How big is the overall project / the share planed for OCR?
General
Speed: What is needed/acceptable? - How much time is available for processing?
Quality: What is needed/acceptable?
Languages: What OCR languages have to be supported?
CPUs used
Exact model and reference (
URL) to specifications, at least the following data:
Image preprocessing
De-skewing
Orientation detection
De-speckling
Restoring perspective
Recognition
OCR/ICR
Languages
Barcodes - what types
Checkmarks
Document structure
Dictionaries
What is the time line for the project?
What technical support terms are required?
Are there any certification requirements?
What kind of documentation is needed?