A few thoughts on this thread:

RAW files - there really needs to be a business case to support this, so 
everyone should start with that - not whether it's better in some respects. 
While I would certainly support maintaining some flavor of RAW (leaning towards 
DNG) in some cases, I'm not sure how well it really applies to facsimile 
collections imaging. While an RGB TIFF with embedded ICC profile is essentially 
a fixed rendition of the object, providing a useful and maintainable image, a 
RAW file requires interpretation which can substantially affect the 
colorimetric and other qualities of the image. Yes, we can store a RAW/DNG with 
a sidecar or other method of storing the RAW processing settings, but in that 
case there's very little to be gained by having the RAW since you're tying it's 
future use to a specific development process anyway.

RAWs are excellent when you *want* the ability to re-interpret the image; 
events images, portraits, etc. For collections imaging, I really fail to see 
the benefit at all, at least as we do our collections imaging here. If we 
wanted to re-interpret the image, we'd then likely have to haul out the object 
for comparison to the image again, at which point we'd be more inclined to just 
re-shoot - afterall, if the RAW processor is better at that point, so (in all 
likelihood) will our cameras be. I fail to see how we're likely to make a 
BETTER rendition of an object in the future, from the same RAW file, without 
having to pull the object back out; and since there is still no standard at all 
between RAW processing software for how to recording the processing settings 
(Adobe uses their own XMP schema, CRS; C1 uses plists; Apple will use something 
else; and so on) and even if there were these settings would be interpreted 
different. Frankly, to me, storing a RAW file for collections imaging would be 
like storing a TIFF without an ICC profile - you'll have little idea how to 
accurately interpret that image in the future without comparing it with the 
object itself.

That said, if there is a case for RAW in your environment, then I'd advocate 
for DNG ideally, and if it fits your workflow today then you might as well 
adopt it right away. If you need to hedge, you can embed the original RAW file 
within the DNG itself, just in case.

Regarding compression and other advanced TIFF features:

Adobe's layered TIFF implementation in Photoshop is ugly last time I looked at 
it, and I would not recommend it as it's unlikely to be supported anywhere. You 
might as well use PSD for in-production files (and if you must have layered 
storage, just ZIP compress your PSD). I have at times uses layered TIFF from 
Photoshop personally, but find it's performance (read/write) is awful and in 
talking with vendors and other imaging geeks think it's highly unlikely that 
particular spec will ever be supported outside of Photoshop, while PSD at least 
has baseline support pretty widely. The only benefit for me of layered TIFF is 
being able to easily write/read metadata from TIFFs using standard tools.

Regarding compression - I see little or no reason to avoid LZW or ZIP 
compression in single-layer TIFFs these days. There are no patent issues I'm 
aware of, which seemed to be the most significant argument against TIFF. Most 
TIFF libraries support RLE, LZW, and ZIP these days. The main difference is 
that LZW is terrible for 16bit per channel data, and ZIP is somewhat slower and 
more memory intensive than LZW with only marginal gains for most 8bit per 
channel images. So I tend to recomment TIFF/ZIP for 16/bpc and TIFF/LZW for 
8/bpc. And of course uncompressed for anyone who has storage capacity to waste 
and can't stand the thought of even lossless compression.

I don't advocate lossy compression for any master files, so JPEG is generally 
out except for access files. I'm quite unimpressed by JPEG2000 today except for 
niche apps, but this could change.

Best,

Roger Howard
Digital Media Specialist
The Getty




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

Reply via email to