Quote:
|
Originally Posted by Steve Jordan
I've wondered myself if anyone else has tried to improve OCR by taking a 2-step scanning process...
|
That, I think, depends on what quality you get 'raw' from the scanner. If the scanner is clunky and produces uneven results in low resolution, it probably would work. I've done it for books printed on bad paper or with uneven press-work.
However, with a reasonably modern scanner, capable of real 300 dpi resolution, and OCR software with the functionality of, say, FineReader 8, you don't need it. You'll need to check thresholding levels (unless you go for greyscale) before you start working, and you may have to check for light levels drifting as the scanner gets warm, but apart from that it's rather plain sailing.
In higher resolution and with good print work, the problem more or less goes away. I've done 600dpi work, and had something like one misread per two pages with only one or two pages of training beforehand.