Re: Colorimetric Accuracy in the Field
Re: Colorimetric Accuracy in the Field
- Subject: Re: Colorimetric Accuracy in the Field
- From: Andrew Rodney <email@hidden>
- Date: Wed, 05 Jun 2013 19:57:48 -0600
On Jun 5, 2013, at 6:43 PM, Henry Davis <email@hidden> wrote:
> Andrew - When I wrote "you have used dE as both a sword and a shield," I figured that readers would understand the inference was that dE is a method for determining a colorimetric match - that dE is pretty much colorimetry itself.
dE is simply a measure of difference. It's quite useful. I use the metric nearly daily. I found it useful to understand how a press output differs due to any number of reasons from a reference. It allows one to see if process control is in effect and if not, how far out it is, and where. It can tell us if two (or more) samples mtatch or don't within limits we define. But how one measures the samples is somewhat important. And without two such samples, the metric is worthless!
The discussion has been about accuracy and dE is part of that discussion but you can't discuss accuracy without those two samples. Someone who says their process is colorimetrically accurate and dosn't measure one of the two sets of samples is blowing smoke up someone's orifice. And measuring two samples of say output of a printer with a Spectrophotometer is a quite different from measuring a sample at the scene (which isn't what one proponent of colorimetrically accurate raw processing has done), since the illuminant in the measuring device is often quite different from the illuminate of the scene! You think that plays a role?
> As an expert in profiling and color management you've insisted that colorimetric matches aren't the purpose of profiling or a color managed workflow.
I've never said that. Or you completely misunderstood what I've said.
> Yet you, and others, offer dE as a measure of the quality of a profile or workflow.
Indeed. Example: One produces a good press profile. One then takes a series of colors, ideally those that fall within that output devices gamut, and sends that data to the printer. One then measures that set of samples and compares it to the reference. Both measurements are made using the same device, same illuminant. A dE value can be used to tell us an average, max, min, Std Dev the two differ. It allows us to know what and where within that color space there is a mismatch. ColorThink and Maxwell are two products that allow us to use this colorimetry to gauge process control and if you know what you're doing, what is going wrong with the press and by how much.
> Thus, colorimetric matches play an important part in evaluating your work - no?
In the example above indeed! This is vastly different from saying a process is colorimetrically accurate without a 2nd set of measurements. Or by dismissing the illuminant that one has to encounter to capure an image. Measuring press sheets, or two different paper samples, or two actual illuminates is quite different.
> You understand the confusion this must cause - I believe that's what you mean by the "rabbit hole".
Measuring two samples with a Spectrophotometer and conjuring up a dE value is simple. Suggesting a colorimetric match without measuring two samples and in this case, dismissing an illuminate a camera is subject to is a deep rabbit hole. And yet, when I suggest the work of Bruce Fraser who produced output from a raw processor that matches the values of a target constitute a colorimetric match, I'm told no, because only 24 colors sync up. Three inches deeper into the rabbit hole. If I measure 24 individual color patches that came off an Epson, compare that to 24 reference values and they are within (you pick the dE), that isn't a valid example of a colorimetric match? Using the same instrument and software to generate the dE report?
> But I don't believe I've heard you give a satisfactory answer in this thread as to why colorimetric matching is and isn't important
Colorimetric matching IS important! Especially once we know the dE limits that constitute an acceptable match and those that don't. Are you familiar with CHROMIX's Maxwell? Very easy to set up what one believes is a match and what isn't. Just enter the max value you will not accept. Send it a sample you measured and it will compare that to a reference, what it calls a Color Aim. In this discussion of accurate color in the field we've not had this value defined. Worse, the process of defining what and how to measure in the field hasn't been defined. That's one part of the rabbit hole. Take Spectrophotometer A and measure two press sheets, hand the Lab (or spectral) values to ColorThink, the answers it provides in dE is simple and non ambiguous. Take a capture in the field. Don't measure even 24 samples in the scene or the scene illuminate, but tell us we have a colorimetric match: That's a very deep rabbit hole.
> You've only offered that it's not important to a majority of photographers.
I don't believe it is. Can you point out where a majority of photogarphers are specially asking for this? Outside the half dozen if that many on THIS list? I hear a lot say their prints don't match their displays. Or over the years, blue's shifting magenta. But demands that their captures be colorimetrically accuate, I'm not hearing this from my peers. They didn't ask for it when using film. Most photographers understood that film had it's own unique color rendering and we picked the film based on our bias.
> You've asked, in a challenging way, for a definition of "colorimetric" from participants in this thread so I figure you're quite ready to spring on any definition with answers that would belittle the pursuit of colorimetric matching. Do you have a definition that's better than the deficient ones in use? I think that might clear up some of the confusion.
I gave you an example above. And yes, I am challenging because the answers are vague and dismiss what appears to be kind of important, namely the effect of the scene illuminate on this process I'm told is Colorimetrically accurate. Or the role of viewer metamerism. Or how silly it is to accept when someone tells you "it matches what I remember I saw". A few others have also asked about portions of what appears to be this rabbit hole. Measuring two printed samples to come up with a colorimetric value is really easy. Some of us have been doing it for 20 years. Bring a cameras to a scene which changes radially, ignore the scene illuminate, or the role a raw processor plays, mix in profiles (DNG or ICC), and the process is quite different and vastly more complex. I never said this matching isn't possible, I've only asked for empirical results that don't dismiss what I think are important considerations in how images are captured with digital cameras.
> May I say that this topic might be interesting to more people than you think, and with your professional experience I bet you can lay it all out in a clear way so that this rabbit hole isn't so dark and mysterious.
It isn't at all complex if you simply accept what some people say without asking the right and sometimes wrong questions. I've tried to do this. But I'm not about to accept that measuring two paper samples with a device is even remotely the same as capturing images and processing them and then pinning the colorimetric match award as being equal. One's a simple process, the other really complex. Hence, that rabbit hole.
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