Re: Accurate color from the camera - who wants it!
Re: Accurate color from the camera - who wants it!
- Subject: Re: Accurate color from the camera - who wants it!
- From: Andrew Rodney <email@hidden>
- Date: Sat, 14 Jul 2007 10:26:24 -0600
- Thread-topic: Accurate color from the camera - who wants it!
On 7/14/07 10:11 AM, "Bob Frost" wrote:
>
> Surely we can measure how accurate or inaccurate our rendered colors are?
Yes we can measure them and that does give us some metric to use when we
talk about accuracy. That's been my point. If you want to say something is
accurate, back it up with measured data. Otherwise its up to interpretation.
That's OK but it causes us to chase our tails too much and often, multiple
users viewing the image just have to agree to disagree.
> Measured the colors of my
> ColorChecker with my i1 (since the published values seem to vary a lot).
> Photographed it with various Nikon cameras. Rendered them with ACR and Nikon
> Capture in different modes. Compared the rendered values with the original
> values and calculated individual and average stats on the closeness of the
> match to the original. That is scientific isn't it?
Yes. I'd say this would fall into a measured and thus accurate scenario. But
now you point that capture device at a totally different scene, one with a
different illuminant, dynamic range and scene gamut and what do you get? Is
that automatically accurate? Does, as some vendors suggest, having a profile
that produces a target that is measurably accurate now mean anything you
capture with this device produce accurate, measured color? I'd say no.
In the case of CR and LR, the technique used with the color checker does
allow some tweaking of the rendering to account for differences in camera
response compared to what Thomas shot using his one or two camera samples in
building his two profiles. He's shooting the 24 patch Macbeth under two
illuminants, now you are too and comparing the results and when necessary,
tweaking the rendering controls to get to a targeted numeric value of the
Macbeth. That's useful but in no way guarantees anything else you shoot will
necessarily be numerically accurate. There lies the rub.
> All of that gives me three LR presets for each camera for skin tone, general
> pastel colors, and saturated colors. Each of those minimises the mismatch of
> the target colors.
Yes, I'd agree. But does that make each scene now accurate? What is
accurate? Will the results always be accurate (once accuracy is properly
defined)?
In Jack's scenario, could one not measure the color of the red sweater,
convert to the numbers to the final working space and tweak sliders to
produce those values in CR? I think so. Does that mean one needs a custom
profile? I think not. Will this tweak produce all other colors in the scene
accurately? Maybe, maybe not. If it did, we'd have little use for selective
color controls in scanner drivers, raw converters and in Photoshop itself.
When you nail the color of one patch on the Macbeth, do all other colors
produce accurate color values? If not, is this an accurate rendering? I
know, a lot of questions. Point is, the term accurate is over used, ill
defined and really wonderful as a marketing term.
Andrew Rodney
http://www.digitaldog.net/
_______________________________________________
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