Re: Profiling Digital Cameras (long)
Re: Profiling Digital Cameras (long)
- Subject: Re: Profiling Digital Cameras (long)
- From: Andrew Rodney <email@hidden>
- Date: Sat, 09 Mar 2002 09:36:06 -0700
on 3/9/02 6:58 AM, Thomas Knoll at email@hidden wrote:
>
There are always going to be some objects that the camera "sees" to
>
be the same color, and the eye sees as different colors. A profile
>
cannot undo this data loss.
Very interesting. The issue's I've found when attempting to profile digital
cameras is more like having a profile that does a pretty good job of
describing the data (not perfect but better than assuming some RGB Working
Space) verses absolute garage. I'm not trying to get a "perfect" profile (if
that's even possible for any device). I simply want something to describe
the data so I can eventually move into a Working Space. What I find is that
if I am successful in producing a profile, I can accomplish that. The
problem is, often I end up tagging the data and getting an awful appearing
posterized looking preview. I'm not sure why this happens as tagging doesn't
alter the data (but the preview in Photoshop is quite accurate to what I'll
end up with once I convert to a Working or Output space even in high bit).
What's really frustrating is that when I attempt to profile a certain
camera, I find I have to throw several packages at the job as some will work
reasonably well and others will not. That is, a package might produce good
results from a certain camera with a certain target and produce awful
results on a different camera system. Throw data from a different camera at
a packages and the opposite will happen. So I have to assume the software
that produces the camera data has some profound effect on the success or
failure of profiling.
The target (standard Macbeth verses DC) seems to be a factor as well
although not as large. I was surprised when Franz at Profile City suggested
I use the standard Macbeth target all the time with their ICC Dcam software.
But he was right, the larger DC target never worked as well.
Last year I tried to profile a Kodak camera (DCS 770). In the past I've had
no problems profiling Kodak cameras when I acquire 10/12 bit linear data
from the Photoshop Acquire module. This time however, the camera was tied to
Kodak's PictureDesk software and NONE of the half dozen software packages
produced anything but awful results.
The fix which worked amazingly well was to use Photoshop 6 and a modified
RGB Working Space as the input profile. Here's what I did:
1. I opened the untagged camera file in Photoshop 6 and it looked pretty
awful (preferred RGB Working Space was ColorMatch RGB).
2. I tried assigning a few different RGB Working Spaces that come with
Photoshop 6. Adobe RGB 1998 helped the preview a great deal but there were
still a lot of problems. The biggest was the preview needed a nice midtone
tweak. I did this by loading Adobe RGB 1998 in the color settings. Then I
clicked on "Custom" and was able to edit the space! I altered the gamma,
tried different chromaticity settings and so forth until I got a pretty good
looking preview. With an untagged file, you can alter all the custom
settings while watching the image update in the background.
3. The "Save RGB" menu saves out this as a new ICC profile. The difference
in the untagged file (using ColorMatch) and my new modified assigned preview
was dramatic. No banding or other issues. The profile did need some
tweaking! The reds were the big issue.
This is where I tried various profile editors. Well not many will edit an
input profile! I had to resort to ColorBlind Edit and was able to do a
selective edit whereby I took a sampling of red and lowered the chroma quite
a bit. From here, the new profile was working VERY well. It was
significantly better than any profile we tried to build even when banding
was the downfall of the profile but not necessarily color.
The idea dawned on me to open the MacBeth ColorChecker in LAB and at the
same time a version that was in RGB and untagged while viewing with my new
custom space. Photoshop does provide RGB and LAB values to the user AS they
are tweaking a custom Working Space as I described above. Unfortunately the
tools provided (White Point, Gamma and X/Y Chromaticity values) is a bit
crude and to tell you the truth, I didn't really know what values to place
in the X/Y fields. But working by eye did allow me to get to a decent
starting point.
The idea of all this rambling is to suggest that there may be a different
way to skin this digital camera profiling issue. Certainly none of the off
the shelf solutions was at all effective and the above technique while not
perfect did solve this photographers color problems. I was pretty happy to
find a solution but unhappy that none of the software I had to produce a
profile worked.
Is the idea of some kind of visual editor (or some way of matching LAB
values of a Macbeth target and one shot under the situation where a profile
is being created) a sound idea?
I have found all the profiling products have produced good results with SOME
cameras and not with others. In my mind, digital camera profiling is still a
black art and not at all a science. A product that makes a very nice profile
for Camera A will make an awful profile for Camera B and vise versa. The
host camera driver seems to have some role here.
Andrew Rodney
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.