[Coco] Scan Test

Roger Merchberger zmerch at 30below.com
Thu Jun 9 14:31:57 EDT 2005

Rumor has it that Michael Wayne Harwood may have mentioned these words:
>Zipping it up doesn't significatly decrease the size - I did attempt it
>and had little success as the PDFs are already fairly well compressed.
>The compressed .tif files from the orginal scans came out to 250mb for 12
>pages at 300ppi/24bit color.

For on-screen viewing (that is our target, right?), you could prolly drop 
the covers & whatnot to 8-bit color with an optimized palette, and most 
(all?) of the inside pages to a 4-bit greyscale... that would ease up a 
*lot* of the storage space right off the bat, yet should still look quite 

Also, the covers could be lossy-compressed, as there is no real need to 
worry about OCR attempts (on the covers) in the future, right?

Running a low-value Gaussian blur on the covers before the palette 
optimization may blend backgrounds for better compression, but may not 
soften edges enough to "worsen" image quality overall...

I do *not* know if you can choose lossless vs. lossy compression in a PDF 
on a page-by-page basis -- that might be up to the PDF rendering suite, or 
it might be document-wide, which IMHO would be bad to choose lossy 
compression then; as that would pretty much kill anyone's chances to OCR it 
later, even for just program listings.

>   For a 150 page issue that would be about
>3gb.  A single layer DVDR holds 4.7gb.  There would be at least 50 DVD's
>needed to archive the high quality images, and while I am going to do that
>for myself I do not envision anyone would want to distribute sets of 50
>DVDs.  The time it would take to burn would be daunting.

I can help with burning -- I have 2 DVD burners available (and could get a 
3rd on short notice) which all run fine in tandem on my home machine -- I 
could make 3 DVDs (at 4x) at one go. SCSI source drive, 1 IDE burner, 1 
Firewire burner (which I've used in tandem successfully) and could get a 
second Firewire burner as needed.

>My experience is that when an image is scanned in at higher quality and a
>separate tool is used to get the images into the appropriate size and
>color depth for the application the results are a better quality than if
>scanned in at lower quality.

Yes. It's easier to optimize too much info, than recreate not enough info. ;-)

>   The 8bit color scans were, in my opinion,
>not as good as the 24bit.  I am planning on doing quite a bit of testing
>and analysis regarding what the best quality/size tradeoffs are going to

I'd offer to help with the scanning except I don't have a decent sheetfed 
scanner. I can help with software testing, however...

Roger "Merch" Merchberger -- SysAdmin, Iceberg Computers
zmerch at 30below.com

What do you do when Life gives you lemons,
and you don't *like* lemonade?????????????

More information about the Coco mailing list