Message-Id: <mailto:199505162156.QAA28772@library.wustl.edu> Date: Tue, 16 May 1995 14:53:39 -0700 From: Peter Rauch <mailto:peterr@VIOLET.BERKELEY.EDU> Subject: Re: Image Database To: Multiple recipients of list IMAGELIB
>Date: Tue, 16 May 1995 14:58:00 EST
>From: Tim Barnett <mailto:0005955798@MCIMAIL.COM>
>library of 500,000 slides and negatives that we wish to digitize.
That's lots of stuff, especially for a "volunteers" and "donations" enterprise.
>We want to stop the clock on the deterioration of the images,
> yet still make the collection available
>to scholars and researchers. The images are used in lectures in
>the form of slides or in video presentations and also for
>producing museum quality prints.
Scholars and researchers find all sorts of information in the detail of an image. See below....
Digitizing is not going to stop the deterioration of the film obviously. You should probably be inquiring about film-to-film copying/archiving and archival-quality storage site conditions, in addition to digital copying.
>We also would like to make the digital surrogate of sufficient
>quality to replace the original in the event of a disaster where
>the originals suffer irreparable damage.
As a digital surrogate (and maybe as the only remaining useful image...), you need to save the "scholarly" detail. See below....
>Given these parameters, can you advise us on what would be the
>ideal 'library resolution' at which we should scan the images?
"Library" resolution is not what you've been suggesting above, but rather "scholarly" quality. I suppose that libraries are repositories of scholarly material (!), but I think their experience with digital, high-detail digital image reproduction is only experimental, with some important studies only now going on to assess "scholarly content" issues of digital imaging.
I would suggest that you explore the following as one criterion for determining how much detail you attempt to digitize/conserve: Inspect a sampling of slides/negatives under a dissecting microscope (i.e., under 10-30x magnification). Have this inspection be done by several scholars of the subject matter apparent in the images (not just the zoologist because it's a picture of a deer, but also the botanist because there are some plants/flowers in the picture, and the geologist because there is a rock formation and a mountain range in the picture, etc...). Ask them to identify fine detail in the images which might be lost if sufficient resolution is not used in the digital copying process. It is surprising (and sad) how much very fine detail is available in many images, that is lost in many of the digitization efforts going on today.
>Also, what about the issue of density range? Many of the images
>are black and white photographs shot in a photo-journalistic
>style in Asia in the 40s and 50s. Detail in the shadow areas of
>these images is important. Should we also set a minimum density
>range for the scanning? (Here I am concerned that one of the
>strong contenders, Photo CD, has a d-max. of only 2.8).
Critical parameters to attend to, for all the "content" reasons suggested by me above.
>One possible strategy that we are considering is first creating
>the database by scanning everything at a relatively low and
>inexpensive resolution. Once everything is catalogued, we would
>then select the more important images and scan those at a higher
>resolution. Do you have any thoughts on this strategy?
The effort to select, batch, clean, orient, insert, remove, and store film for your low and "inexpensive" resolution images might well be the most costly part of the work. It is all labor-intensive. The actual digitization is usually a small proportion of the time/effort involved, and the cost of storing the more voluminous high-res images, while significantly more than for the low-res images, is still probably a modest proportion of the overall cost of digitization. You can always extract low-res images (onto higher-cost storage media) from the hi-res ones (stored on lower-cost storate media), for everyday use.
Photo-CD is worth looking into, as its resolution is more or less adequate for much of the "scholarly content" detail I allude to above. Color/density issues need to be experimented with. Given a collection of 500,000 film images, you need to do some representatively valid sampling to determine what "populations" of characteristics these images contain, e.g., 30% "blue" off-color images, 10% low-detail images, 60% very high detail images, 20% very dirty/dusty images, physical locations of these "populations", ...., so that you can consider strategies for batching the materials to process like materials together, for greater efficiency/consistency/quality and less cost. (You can always rearrange the digital material later; you don't have to digitize like-subject matter together, if other considerations argue for more economical digitizing by some other criteria.)
You didn't mention the size/format of the negatives, but even if they are larger than 35mm, they might also be digitized with PhotoCD. Peter