Title: Re: Digital Imaging Best Practices - Resolution
Hi everybody,

I couldn't agree more with Peter - just because you crank up your resolution (which results in a bigger filesize) doesn't mean you've actually captured more data worth having...

I'd like to throw some light on this issue from a slightly different angle and address the issue that most discussions around direct capture / scanning revolve exclusively around resolution, which in and of itself is fairly meaningless as a measure of whether an adequate amount of data has been captured.

Just to make that point clear with some examples: If you specify data captured by pixels per inch (ppi), and you capture a very small object, you wind up with a file that's unsatisfactory for most uses. Let's say our original material is the size of a slide (1.5 x 1 inch), a 600ppi capture standard results in a file of 900 x 600 pixels (1.5 MB). If our original material is the size of a handscroll (e.g. 12 x 60 inches), a 600 ppi capture standard results in a file size of 7200 x 36000 pixels (740 MB). In the first case, a ppi standard results in a file so small it's almost useless, in the second case it demands an impossible capture (exceeds hardware capabilities).

An alternative way of specifying data capture is to standardize on pixels per longest dimension. A middle-of-the-road standard would be 3000 pixels / longest dimension - that spec fits the file a 4x5 transparency yields at 600ppi, so it seems like a good equivalent to a 600ppi standard. For our small item, that standard would result in a file of 3000 x 2000 pixels (17MB), and for our large item in a file of 3000 x 600 pixels (5 MB). This way of specifying data deals better with our small item, but it doesn't deal well with items that have odd proportions - our handscroll only yields a fairly small file.

As you can see, one of the other measures I'm always using to judge whether I think my file is good enough in terms of sheer numbers is the file size in MB - an absolute figure giving you the amount of raw data. The MB figure however does not take into account the size of the original object - if you have a rather big object, even a fairly high MB count wouldn't mean that your file has captured enough detail of the material at hand to be satisfactory to you. In the end, it's useful to weigh all those factors, and establish a range of tolerance. You may want to calculate both the MB file size for a ppi standardized capture and for a capture standardizing on a longest dimension, and then weigh the outcomes against what's practically possible. You could then go ahead and adopt the capture which yields the greater amount of data.

As far as absolute file size in MB is concerned, at BAM/PFA our direct captures usually yield between 60 - 120 MB, which means we capture in a range of roughly 6000 pixels / longest dim or roughly 1200 ppi.

All of the calculations above assume an uncompressed file at 24 bit. You can play around with figures using the following formulas:

Formula for determining filesize (uncompressed):
File Size (k)= (height inches x width inches x ppi x ppi x bit depth) / 8192 (number of bits in a k)
or
File Size (k)= (pixel dimension x pixel dimension x bit depth) / 8192
(further divide by 1024 for MB)
Bit depth = RGB 24 (8 bits per channel)

Another thing to keep in mind when capturing from film surrogates instead of original material is the actual data contained in the film stock. Here are some rough figures of how much information is contained in a 35mm slide and a 4x5 transparency. These figures assume that your photography is absolutely pristine (and most of course isn't.)

35mm ~ 4000ppi ~ 60 MB
4 x 5 transparency ~ 2000ppi ~ 230 MB

Hope this helps a little bit to make a thoroughly confounding topic a little more transparent...

Cheers,
Guenter


At 6:01 PM -0400 9/12/02, Peter E Siegel wrote:
Hi all,

Useful resolution is a complicated subject, to just say making an image
with more pixels is not always going to yield better pictures. You must
consider the optical path, the sensor itself (Beyer type, monochrome,
trilinear etc), the mechanism that moves the sensor, the source material,
and generally what you want to use this image for. If it is a replacement
image of an existing film or photograph, it is very hard for even drum
scanners to get all the useful data; you might need a photo-densitometer
to get everything. After one decides on the longevity of the digital
master, one needs to make sure to make good pixels, not just more. Then
there is a trickier issue about color... then it gets worse.


My 2 cents,


Peter


On Thu, 12 Sep 2002, Richard Urban wrote:

> Hi all,
>
> As part of the Western Trails Project, the Colorado Digitization Project is
> reviewing our best practices for digtial imaging that were established in 1999
> (see http://coloradodigital.coalliance.org/scanning.html).
>
> One question we are wrangling with is whether to increase the minimum
> resolutions for digital images.
>
> A review of a number of other digital project  websites shows that our current
> guidelines are still comprable to our peers, although many of them have not
> been updated since ours were established.
>
> We'd like to hear from other projects that are exceeding the current baseline
> targets to talk about the pros and cons of higher resolutions.
>
> Thanks,
>
> Richard Urban
> Operations Coordinator
> Colorado Digitzation Project
> [email protected]
> http://coloradodigital.coalliance.org
>
>
>
> ---
> You are currently subscribed to mcn_mcn-l as: [email protected]
> To unsubscribe send a blank email to [email protected]
>

___________________________

Peter Siegel
Head, Digital Imaging and Photography
Harvard University Art Museums
tel.    1-617-495-4846
fax.  1-617-384-8203
email.  [email protected]
addr.   32 Quincy Street
        Cambridge, MA 02138
        USA


---
You are currently subscribed to mcn_mcn-l as: [email protected]
To unsubscribe send a blank email to [email protected]


-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Guenter Waibel
Berkeley Art Museum & Pacific Film Archive
Digital Media Developer http://www.bampfa.berkeley.edu/
Digital Media SIG Chair, MCN http://www.mcn.edu/SIGS/Digital/visig_subscribe.htm
[email protected]
Phone     510-643-8655
Fax 510-642-4889
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

---

You are currently subscribed to mcn_mcn-l as: [email protected]
To unsubscribe send a blank email to [email protected]

Reply via email to