Re: Untagged RGB to Lab conversion
Re: Untagged RGB to Lab conversion
- Subject: Re: Untagged RGB to Lab conversion
- From: Robin Myers <email@hidden>
- Date: Wed, 19 Dec 2001 13:09:31 -0800
- Organization: Robin Myers Imaging
David Remington wrote:
>
... I'm curious, and a bit concerned, about what
>
has happened before you even open the file. I've worked with and
>
evaluated many cameras at this point and find their raw, by raw I
>
mean gray balanced and camera processed (exposure and contrast
>
referenced to a gray scale), data more or less mismatched to standard
>
editing spaces. In general the image is oversaturated in any of the
>
large gamut spaces. This was most noticeable with a Jenoptik camera I
>
looked at. That data fit Colormatch most closely. So, to clarify,
>
does in camera processing include targeting a particular
>
(anticipated?) color space? How does this effect the integrity of the
>
capture? Is the data clipped and or de-saturated? There must be some
>
strategy for the selection of RGB values.
Firstly, there is no "standard editing space". The editing color spaces of Photoshop are
unique to Photoshop, and, in my opinion, not necessary. They only have usefulness in
certain workflow situations. If your image is tagged with a profile at generation, then
an editing space is unnecessary.
The only cameras I know of that put their images into a particular space are the consumer
cameras that provide their images in sRGB. The image is highly processed and corresponds
to a perceptual rendering of the original scene. The capture integrity is fine, it is the
post capture processing that can make or break an image. This post-processing depends on
the camera manufacturer, user selected exposure options, the exposure itself, ...
>
>> Is RGB data completely defined by the box you draw around it?
>
>
>
>I do not understand what you mean here? Please clarify.
>
>
I guess this is a reiteration of RGB device dependence. Is it true
>
that without a color space (primaries, white point, gamma, or LUT)
>
RGB values are meaningless? I'm just wondering how to best evaluate
>
the accuaracy and integrity of a digital capture and have hit a snag.
>
I can't fairly judge the file in my standard editing space and I
>
can't convert to Lab. What is left?
Without knowledge of the primaries, white point and tonal reproduction curve (i.e. gamma
for monitors) the RGB values are less meaningful. They might look good on a monitor, but
it is because the RGB responses of the camera were reasonable matches to the RGB monitor
emissions.
Color is only one attribute you should evaluate. Image noise, blooming, the ability to
capture and maintain highlight and shadow detail, number of stops of captured lightness
range, exposure adjustment capability, etc. are just as important. If you must compare
color capabilities, use a calibrated monitor and place side-by-side images from the
different cameras at their best possible quality and compare them visually to a match
with your ColorChecker. If you need to compare L*a*b* values, you might use the monitor
profile as your camera profile and convert to L*a*b* (there are reasons for this that I
will explore in a white paper soon).
Sincerely,
Robin Myers