Scanning "Results"

Don THX1138 at
Sun Sep 3 10:26:58 CDT 2006


I spent a few hours last night playing with different
scanning options and figured I would share my observations.
Definitely not a "scientific experiment" but, rather, "just

Image in question was just a page of text -- probably 8-10pt.
Laid out in two columns, quite a bit of whitespace.  The
original image dimensions are about 8.5x12" (yes, 12, not 11).

[Sorry, in retrospect I should have used an 8.5x11 image as
this would be easier for most folks to relate to  :<  Instead,
I just set the scanner to scan half the available area (it's
a B-size scanner)]

First, I did a monochrome scan at 400dpi (which is where I
tend to do most of my scans).  The resulting TIFF file was 2MB.
When viewed "on screen", the TIFF file (i.e., eliminating any
effects of the scanner software) was quite readable.  No signs
of jaggies, etc.

I ran that image through various compressors (still sticking with
TIFF).  "Packed bits" yielded a file size of 360K -- as did
Huffman encoding.  LZW dropped this to 220K.  CCITT3-1D encoding
dropped it to 217K while CCITT3-2D brought it down to 131K.
And, CCITT4 brought it down to 100K.

I then ran the same image at 800dpi (what the heck, let's live
dangerously!!).  As expected, the original TIFF grew to 8M.
The CCITT4 variant grew to 250K.  (I didn't bother with all
the other encodings as these two represent the aparent
extremes of monochrome representations *EASILY* AVAILABLE TO ME).

Next, I scanned the same image at 400dpi in *color*.
A 24 bit TIFF took a whopping 55MB.  (can you say, "Sorry,
but we don't got no bananas...").

I then tried to save the image as a JPEG -- *guessing* at
appropriate compression/smoothing factors to get the resulting
image size down to ~1MB.  (for reference, ASSUMING THESE
"settings" ARE PORTABLE, 4:2:2, 77 compression, 10 smoothing,
optimized but NOT progressive).  I got lucky (?) and this
first pass gave me a 530K image.

With the 250KB *800*dpi B&W C4 TIFF in mind, I decided to
push the file size even smaller.  (compression increased to 90)
This resulted in a 360K image.  Even more squeezing (compr = 95)
got this down to 250K.

However, the *quality* of the image was very disappointing!
The 530KB version was quite "fuzzy" (not "jaggies", since JPEGs
are more continuous tone than a B&W TIFF, but, rather, "blurry").
The 360KB version began introducing noticeable artifacts that
were clearly not present in the original image.  This got
worse in the 250K version.

Bottom line:  the 100KB B&W TIFF was much better looking
than even the 530KB JPEG.  And the 250KB B&W TIFF was so
"fine" that I suspect it is overkill (I doubt anyone or
anyTHING -- i.e. software -- could discriminate between
that at 800dpi and the 400dpi version).

I had earlier tried some gr[ea]y scale scans and convinced
myself I must be doing something wrong (as the sizes were
just so much larger) so I didn't pursue them.  In hindsight,
a more scientific approach would try to JPEG encode those
monochrome representations as well.  (I'm through experimenting
as I have the answer *I* want/need)  Somehow, I doubt they
will prove to be more economical than the C4 TIFFs.

[N.B. the gr[ea]yscale scans are much "softer" on the eyes
(no doubt due to the continuum of "value")]

So, to answer *my* initial question (ages ago), 400dpi C4 TIFFs
are definitely "adequate".  800dpi are overkill.  And, at
~100KB per page, they are quite "affordable".

We now return you to your regularly scheduled program...

More information about the cctech mailing list