Re: 16 bits = 15 bits in Photoshop?
Re: 16 bits = 15 bits in Photoshop?
- Subject: Re: 16 bits = 15 bits in Photoshop?
- From: email@hidden
- Date: Mon, 18 Apr 2005 13:51:13 EDT
Bruce Fraser writes,
>>One of the hidden variables in Dan's "challenge" is how much of that work
has already been done in high-bit space. If you start with a linear, uninverted
scan of a color neg, for example, there's simply no contest between the 8-bit
and 16-bit versions.>>
There is plainly no point in examining your religious tenets further, but I
must correct a factual error in the above.
I have done *exactly* that kind of testing, to around a dozen images--both
positive and negative film, both film and digital captures.
A linear, uninverted 16-bit scan of a color negative, for example, was the
basis of Figure 15.11 of Professional Photoshop Fourth Edition and the original
data is on the book's CD for those wishing to verify. The original image was
almost monochromatic and so flat that it was difficult to determine the subject
matter. I applied absolutely massive corrections to bring it to
acceptability, once in 16-bit, once in 8-bit. The book shows the two corrections at their
intended size, quarter page, and then compares two sections of the image at
400% magnification, showing both composite and magenta channels.
A jury shown the Matchprints blind chose the 8-bit version as slightly better
than the 16-bit. There was no registration problem with the Matchprints.
>>But nobody in their right mind would do such a thing except to prove a
point.>>
Agreed. That particular image was done to prove a point. The edits were so
extreme that one could argue that if that picture didn't show an advantage, then
it's hard to imagine one that could. Also, anybody whose workflow features
edits of that magnitude has far more significant things to worry about than bit
depth.
Similarly, a professional photographer late last year took two of his raw
digital images, both of subjects with gradient areas that might be expected to
band, and, to prove a point, exported the linear data directly into Photoshop,
and opened in Adobe RGB, resulting in images that were grossly, grossly too
dark. He then corrected them, once in 16-bit, once in 8-bit, but to prove a
point, he did so in a clumsy and amateurish way where certain of the corrections
worked against one another, although the final images were acceptable. His
8-bit versions showed, as he had expected, unpleasant noise. They could be
characterized as a "night and day", "totally obvious to anyone who looks" kind of
difference. He shipped me all his files and backup materials, and I redid his
tests and got the same results.
On examination, it was discovered that the 8-bit and 16-bit files were
generated not in Photoshop, but by his Canon software. The 8-bit file was defective,
being found to be typically 2.5 levels darker in most areas than the 16-bit
file. I then made a copy of the original 16-bit, converted it to 8-bit in
Photoshop, and reapplied all the corrections.
The result was, no significant difference between the 8- and 16-bit versions.
The photographer, whom I do not know, has confirmed publicly that he agrees.
Before doing any such testing, I require that the photographer give me
permission to publish the results; I have that, and so the stuff will presumably
appear in Professional Photoshop Fifth Edition--unless I get something better in
the interim.
So, the point was proven. "No contest" is eminently contestable.
Dan Margulis
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden