[Coco] Re: Feedback on Quality Requested

[James Hrubik]

Gene, I think the terms of Michael's license require exact replicas;  
OCR is not an option IIRC.  Michael can address that, though.

Michael, I never did hear back whether the quality of the final issue  
I posted for you was OK.  I extracted page 1 from it and it will be  
up on

http://www.acorn.net/aacc/sample.pdf

for a short time for everyone to comment on.  It is a .pdf scanned at  
300 dpi in B&W, then diddled in GraphicConverter to add the color  
back in, and finally .pdf-ed with Acrobat.  Very labor intensive.   
The image quality is, I think, close to as good as can be gotten from  
the old yellowed newsprint, but then again, it was scanned on my old  
scanner, and the new one has more options, as well as being LOTS  
faster.  In source .tifs, each color page was well over 1 meg, but  
Adobe crunched it nicely.  The whole 16 pages is a 9 meg file.  At  
that rate, a 200 page issue would be somewhere in the neighborhood of  
75 megs (pages in the mag issues are smaller than in the newsprint  
issues).  Adobe builds  in a lot of overhead; I extracted page 1 for  
the sample, and it was 1.8 megs for just the one page.  Ridiculous.

I scanned an out-of-print book, over 300 pages, for my grandkids, at  
600 dpi in 8-bit greyscale.  The quality is fabulous, but the .pdf is  
over 195 megs.  I think there is a point where we will have to  
compromise on quality to get the best fit on the disk.  I agree with  
Gene, though, the original scans should be burned to multiple archive  
disks; the future will hold better compression tools.

On Apr 27, 2006, at 8:07 PM, Gene Heskett wrote:

> On Thursday 27 April 2006 19:25, Michael Wayne Harwood wrote:
>> Richard,
>>
>>> Since the project will be sold at about $60-$70 dollars, seems to me
>>> the main issue should be what will be the quality of the magazines
>>> to be scanned. Based on the djvu demo, the quality of the magazines
>>> will be more important than the processes used to acquire and store
>>> the images. Paying $60 for images of damaged magazines is food for
>>> thought.
>>
>> Bear in mind the number of pages is close to 25,000 so for a single
>> 4.7gb DVD to include everything the average size per image will need
>> to be around 150kb to 170kb depending on how much space will be
>> needed by the "Rainbow on Tape/Disk" data files and the searchable
>> index.
>
> My own opinion is that shooting for a sub 200k size per page is  
> probably
> going to sacrifice quality in many ways that would come back to haunt
> us later.  So I would, for the src archives at least, tend to want
> pages of at least 10 megs so that they could be downconverted the
> minimum amount to fit the media, which by then might be HD or blue- 
> ray,
> with 30-50GB of capacity per disk.
>
> How much storage area do you have available for this project?  And can
> you do OCR on the pure text pages, then output a 10k .ps file for  
> those
> pages only.  I've found that reliable OCR that doesn't need a lot of
> hand editing, tends to need 600dpi and up scanned image to work with,
> and those will be 20+ megs a page until converted.
>
>>
>> Regards,
>> Michael Harwood
>
> -- 
> Cheers, Gene
>

---------------------------------------------------
-----Items below rated "R"; parental discretion advised----
---------------------------------------------------
"Brilliant minds, like productive gardens, flourish under the  
influence of bullshit."
---------------------------------------------------
  From the sayings of Grampa Jim, Copyright 2006.
Unauthorized use of my stuff may cause senility.