Scanning Formats

Don THX1138 at
Tue Aug 15 11:43:57 CDT 2006

Barry Watzman wrote:
> RE: 

[fixed indents]

>>> But doesn't JPEG use lossy compression?
>> Yes.  You can adjust the "quality factor" but I think this
>> would be A Bad Choice.  Especially given my original comment
>> that the documents are D-size reduced to B-size before printing
>> (i.e., the text/lines are VERY fine and apt to disappear
>> in the DCT application within the JPEG encoder.)"
> Yes, it's lossy compression, but if you use a high-quality setting (which
> was my point of 5K to 10K file size per square inch for color, or about 1/3
> that for monochrome), you will never notice ANY difference.  The size of the
> document is irrelevant, since as I said, "5K to 10K ***PER SQUARE INCH***".

The size of the *original* document is important as it has been
scaled down to print on a page 1/4 it's original size.  I.e.
you start to approach the resolution of the marking engine
(e.g., 1 pixel wide lines) which, mathematically, looks like
a very high frequency component.

> Larger documents produce larger files, but the quality loss for larger
> documents is no greater than for smaller documents ... which is to say,
> imperceptible.  TRY IT, and compare the results.  I suggest that you will
> not notice ANY difference between JPEG and TIFF, no matter what you do, no
> matter how you examine them.  I have scanned nearly 40,000 pages now, plus
> almost 10,000 color photographs, since 2001.  There is no good reason to use
> TIFF instead of JPEG, as long as you don't try to achieve excessive
> compression levels with JPEG (and at 5K to 10K per square inch for color,
> you won't be doing that -- that equates to 500K to 1 megabyte for an 8.5" x
> 11" color page).  In fact, I often find that I can go down to less than 100k
> bytes per page (that's down towards and below 1K per square inch), but this
> is too much compression for some documents.

I'll play with it "as an exercise".  I suspect I will find
fuzzy signal lines (i.e. as if they were anti-aliased)  :-(

Let me dig up some suitable drawings (any B-sized photocopies
will have too much "toner spread" to test the fine line issue)

More information about the cctalk mailing list