djserian@LAKEHEAD.BITNET (02/05/91)
In responce to a posting of mine regarding the Scan Man Plus
beaumont@zoo.sdd.trw.com writes:
I bought an original ScanMan several years ago. I was happy with it at
first, using it to digitize photos. But it gradually got worse and worse
with its scanning. Finally the scans got so bad that I complained to
Logitech. They had me scan a picture and send them the output and the
original picture. I figured they'd be able to tell me what was wrong with
the thing so I could get it fixed. Wrong!
They sent me back my picture and the output scanned with a ScanMan Plus
(which looked pretty good) -- along with an offer to upgrade my ScanMan
for "only" $150 US. I debated awhile and finally went for it.
The Plus is orders of magnatude better than the original ScanMan: Rollers
on all four corners, adjustable dithering, and 100-400 DPI scan densities.
The yellow-green filter got rid of the annoying red-sensitivity the old
ScanMan had -- and added an annoying blue-sensitivity. [Scan a photo
with a blue background and it will come out black.]
Paintshow Plus is (barely) adequate for editing the pictures, but it is
limited by the available DOS memory. After dumping my TSRs, I can get
a picture that's about 1.7 megabits in size, which is pretty small at
400 DPI. The latest version of PSPlus has an annoying and confusing
"feature" which tries to resize the picture based on the printer aspect
ratio, the screen aspect ratio, and the scanner's aspect ratio. There's
apparently no way to make it default to something sensible, so I have
to remember to reset it at the beginning of each session. The most
annoying feature is perhaps the difference between the scanner's dither
patterns and the tile size: The scanner dithers in multiples of 3 pels,
and the tile size is in multiples of 4 (16, actually). Editing scanned
pictures is reasonably painful, since you can't "paint" with one of
the scanner's dither patterns (won't line up). It can be done, however,
with a lot of patience.
I have found one major bug in PSPlus: When running in a high-resolution
mode (640x480x1 VGA, for example) it apparently miscalculates the amount
of memory available for the cut/paste buffer. If you try to cut an area
of the screen that's too large, the bottom N lines get erased from both
the screen and the cut buffer. Gone forever.
As for Catchword, my recommendation would be to save your money. I bought
a copy to scan in some standards documentation and found that it took MUCH
longer to scan in a page than it did to type it in! You can scan a page
in either vertical or horizontal strips. When scanning vertically, the
program is supposed to "stitch" the two strips together, but this doesn't
work very well. You have to overlap the scans by 1/4" and you have to
scan within 3 degrees of vertical. I carefully scanned in one page, and
when it couldn't stitch together a couple of paragraphs it just erased
them from the input!
After you scan some text, you have to wait while it builds its pattern
data base and then tries to match up the patterns with letters. On my
10MHz 286, this takes about 10 minutes (even with 1.4MB of disk cache).
Then you have to "help" the recognizer with things it couldn't figure out,
another 15-30 minutes PER PAGE. After that, you get to edit the document
and fix things that it did wrong, perhaps another 10-20 minutes. An hour
a page isn't exactly speedy in my book. I could type the whole page with
fewer errors in 10 minutes.
CatchWord claims to regognize a lot of different fonts and sizes, and it
may indeed do that "well", based on state-of-the-art and cost. It might
work OK with typewritten documents, but it does poorly with kerned text.
Letters which "tuck under" each other (like "WA" or "AV") tend to become
one amorphous blob. It recognizes serifed fonts, but doesn't do well if
the ascenders are light. The "W" in the font I was using has very wispy
ascenders, so it kept scanning Ws as "\\'". Blurry text, such as something
that's been xeroxed a number of times, doesn't scan at all.
CatchWord claims to handle underlined text (it doesn't) and to eliminate
boxes around text and other art (it doesn't). I suppose it would work
better if it was fed data from a full-page scanner which would eliminate
the human "shakey-hand" input, but I'm not in a financial position to
try.
***********************************************************************
* Derek J.P.K. Serianni * E-Mail: djserian@flash.lakeheadu.ca *
* * Phone : (807) 344-5170 *
* Computer Advisor * *
* Lakehead University * Insanity is just a state of mind! *
* * *
* Thunder Bay, Ontario * Computer problem? - NO PROBLEM! *
* CANADA * *
***********************************************************************
QUIT