Re: Document Scanner


I'm not sure what you mean by "Kurzweil can otimize the scanning configurations..."  If you are talking about the feature you can run that takes a sample of the page and uses what it thinks are the best settings, I did a few tests with it years ago and consistently found that just using the defaults is better.  And when scanning a book, often, in my experience, a page or part of a page will look different, presumably the contrast changes for some reason.  I therefore always, unless results are poor, use whatever the setting is for automatically setting contrast as you scan.  I don't recall what either Openbook or Kurzweil calls this setting.  As far as I know, the setting is the default in both programs.  I've looked at settings very infrequently for years since I did a good deal of experimentation to see which work the best. 
As for Openbook's having the image recognized by more than one OCR engine, I haven't played with it much but the little that I have, I have found the fastest setting to be more accurate than the slower and claimed to be more accurate settings.  My guess is that fastest only uses one OCR engine and slower settings use more than one.  Of course, since company after company dumbs down there products names for things, you can only infer from the performance.  In the old days, you could actually tell Open book what OCR engine or combination to use when that feature became available.  Not any longer.  Now generic meaningless descriptions are used such as slowest, best quality.  Don't assume that any such descriptions are correct.  Test and see.  In my small amount of testing with books, I find the Openbook descriptions to be inaccurate.  That's not to say that there are no occasions when they describe what happens in terms of accuracy, I don't know.  I'm only saying that in what I would say are more or less typical scanning conditions, scanning a few test pages from different books that are probably of typical scanning difficulty, I invariably found the fastest setting which is claimed to be least accurate to be most accurate.  Perhaps this is another illustration of the saying "Too many cooks spoil the broth." 
I question whether Kurzweil and Openbook provide better recognition than Fine Reader.  I don't know how current versions of Fine Reader are set up but the version I got bundled many years ago with my scanner, had a setting for what I'm describing as automatic contrast.  Not the setting where it scans part of a page, then infers the best settings to use and then scans the entire page, but a different setting.  Again, as with Kurzweil, I found this setting to work much better than the setting I described with part of a page being scanned each time and scanning settings being inferred.  That is the rough equivalent of the Kurzweil setting I described at the start of my message.

----- Original Message -----
From: Rui Fontes
Sent: Sunday, July 30, 2017 9:06 AM
Subject: Re: [nvda] Document Scanner

Both produce better results, due to some enhancements and configurations
Kurzweil can otimize the scanning configurations and Openbook can use the
best of each OCR's to provide a otimal result...

But, forgeting all technical stuff, both provide a interface very easy to
use, allowing everybody to press a key and start listening the document...


-----Mensagem Original-----
De: Rob
Data: 30 de julho de 2017 11:18
Assunto: Re: [nvda] Document Scanner

Gene <gsasner@...> wrote:
> I'm very skeptical that Kurzweil; or Openbook provide either better
> recognition or recognition that is better enough to pay those exhorbitant
> prices for.

Kurzweil uses fine reader as its OCR engine, anyway. What you gain is better
fine tuning of the OCR process. With the commercial version of FR, this is
done by drawing blocks around the image portion, highlighting and enhancing.
In Kurzweil, you can do this by using its optimize scanning features, which
can be useful for old documents and blurry print.


Join to automatically receive all group messages.