AmigaActive (308/1728)

From:Kevin Twyman
Date:6 Jun 2001 at 19:21:25
Subject:Migraph OCR

Hello Don,

On 06-Jun-01, you wrote:

>> I found it again recently, and had another play with it. It's
>> amazing what a scanner and 060/50 does to that program, it loads,
>> it OCRs, it even learns, and it does it at a very acceptable speed
>> as well.

>> Obviously a product well before its time :-/

> I use it now and then. It can work quite welll if the original is
> clearly printed type scanned at about 300 dpi. If the original has
> broken or joined-together letters, forget it.

That's what I thought, until I persevered and used the learning mode
to teach it to recognise such flaws in a text. Out of three A4 pages,
the first was quite slow, the second much faster, and the third went
right through without a problem.

> The main thing is to realise that they use the word "dictionary" to
> mean "typeface".

This is where I was going wrong. For each typeface MG works best with
a custom dictionary. In practise I've found that it will find several
typface faults incorporated in the same dictionary, and the one I've
built up will happily convert most scans from books, magazines etc.

The other thing is to get a good monochrome scan, in TIFF or IFF-ILBM.

> I think the best Mac and PC programs are far ahead here.

Agreed.

> We could certainly use a good modern OCR program for the AmigaOne.
> Somebody did try to write (or port) one a couple of years ago, but
> it was useless.

It's an area that I don't think many people venture into, but it can
be dead handy. My latest project was a photocopied extract from an old
article about past times, told by a relative some years ago. I have
now been able to extract the text and incorporate it into my Family
History archive, with very little trouble.

Regards,



Kevin - Est. 1949 http://www.theboyz.freeserve.co.uk
A man with an Amiga A4000T CyberStormPPC 604e/233

Whaddya mean you don't STAPLE diskette labels on?

Quote carefully and read all ADMIN:README mails

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/