From: "David Nissen", INTERNET:dnissen@home.com
Date: Fri, Jul 20, 2001, 9:37 PM
RE: [colortheory] Why no red-cyan channel in LAB?
In designing LAB, after choosing luminosity as one dimension, 2 dimensions were left.The designers chose green-magenta, the most important (?) of the primaries, and blue-yellow the least important (?). I think this is true in terms of the weighting in luminosity.
Red-cyan was not included.
Is there some rationale for this particular 2 out of 3?
Just wondering.
Thanks,
David Nissen
dnissen@home.com
From: Dan Margulis, INTERNET:76270.1033@compuserve.com
Date: Sat, Jul 21, 2001, 4:16 PM
RE: [colortheory] Why no red-cyan channel in LAB?
David writes,>>In designing LAB, after choosing luminosity as one dimension, 2 dimensions were left. The designers chose green-magenta, the most important (?) of the primaries, and blue-yellow the least important (?). I think this is true in terms of the weighting in luminosity.>>
The A and B channels don't have a luminosity component, but for all real-world purposes, you're right--the A channel is vastly more important than the B. This is one of the reasons that profiles generated by a series of machine measurements suck so often and so badly, but that's another thread.
>>Red-cyan was not included. Is there some rationale for this particular 2 out of 3?>>
In some ways it might be better for us had red-cyan been used rather than blue-yellow. However, the human visual apparatus is strongly suspected of doing its proprietary color evaluation along the magenta-green and blue-yellow axes as opposed to some more obvious way. For example, in the most common form of color-blindness, the affected person can distinguish yellow from blue to some extent, but is completely unable to tell magenta from green.
Therefore, so the rationale goes, if one set up a red-cyan channel, it might not be possible to encode it so that increases or decreases in value would appear smooth and consistent to the human observer. With yellow-blue, the idea is that a 10-point change in value would create approximately the same impression of a change in color no matter where on the curve it fell.
Dan Margulis
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Sun, Jul 22, 2001, 3:24 AM
RE: [colortheory] reasons why profiles suck so bad
Dan writes:
>The A and B channels don't have a luminosity component, but for all
>real-world purposes, you're right--the A channel is vastly more important
>than the B. This is one of the reasons that profiles generated by a series
>of machine measurements suck so often and so badly, but that's another
>thread.If it's another thread then you probably shouldn't bring it up in the first place. So long as the device actually behaves, and continues to behave per the measurements forming the basis of a profile, they work remarkably well. I wouldn't say they "suck so often and so badly." That's a complete misrepresentation of the technology.
They certainly aren't perfect by themselves but that has less to do with Lab and measurements than it does screening algorithms and the fact *all* color space conversions are performed on pixels. Even if we talk about ink settings files, or other proprietary forms of necessary color space conversions they all occur on pixels and do not take the effect of screening into account.
The only reason I'm aware of as to why a profile would "suck so often and so badly" is because the device behavior is inconsistent, and therefore a profile can't be made quickly enough to adequately describe a device's behavior. Maybe you'd like to be a little more clear on what you're trying to say?
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932
From: Andrew Rodney, INTERNET:andrew@digitaldog.net
Date: Sun, Jul 22, 2001, 11:43 AM
RE: Re: [colortheory] reasons why profiles suck so bad
on 7/22/01 1:24 AM, Chris Murphy at lists@colorremedies.com wrote:> If it's another thread then you probably shouldn't bring it up in the
> first place. So long as the device actually behaves, and continues to
> behave per the measurements forming the basis of a profile, they work
> remarkably well. I wouldn't say they "suck so often and so badly." That's
> a complete misrepresentation of the technology.
>Ain1t that the truth! Again, one could ask Dan about his experiences with profiles (what packages he1s used, how he comes to such a broad conclusion) but it usually falls on deaf ears. Dan1s assumptions about how profiles behave (they suck) is like saying one group of people (be it African Americans, Italians etc) behave a certain way. It1s a prejudice that has no backing in rationality. Just like most prejudices. Yes, some profiles suck. Some scanners, cameras and Photoshop operators suck too. That isn1t at all an indication of the majority at large. Time for a reality check Dan.
Andrew Rodney
From: "Ron Bean", INTERNET:rbean@execpc.com
Date: Sun, Jul 22, 2001, 2:48 PM
RE: Re: [colortheory] reasons why profiles suck so bad
Andrew Rodney writes:>Dan1s assumptions about how profiles
>behave (they suck)...Hmm, I thought he was specifically talking about
machine-generated profiles. One of us missed something...>Time for a reality check Dan.
OK, here's one:
>This is one of the reasons that profiles generated by a series
>of machine measurements suck so often and so badly...What does "generated by a series of machine measurements" mean to you? (Or more to the point, what does it mean to Dan?)
Weren't you telling me a while back that profiles have to be tweaked by hand? (Or maybe that was Chris...) In any case, how about a little less heat and a little more light...
From: Andrew Rodney, INTERNET:andrew@digitaldog.net
Date: Sun, Jul 22, 2001, 3:12 PM
RE: Re: [colortheory] reasons why profiles suck so bad
on 7/22/01 12:47 PM, Ron Bean at rbean@execpc.com wrote:> Hmm, I thought he was specifically talking about
> machine-generated profiles. One of us missed something...What do you mean, "Machine-Generated profiles"? You can't create a profile without a "machine (a Spectrophotometer) unless you want to talk about the low end products that use a scanner instead. That's a "machine" too and for some devices, a scanner can make good profiles and in others the profiles suck. I know of no package that makes output profiles with your eye!
> What does "generated by a series of machine measurements" mean to you?
> (Or more to the point, what does it mean to Dan?)Good question!
> Weren't you telling me a while back that profiles have to be
> tweaked by hand? (Or maybe that was Chris...) In any case, how
> about a little less heat and a little more light...Most profiles made with a good package and good hardware don't need tweaking. For cross rendering (make one printer match a completely different printer like getting an Epson to match an Iris), you usually DO need to tweak the profile. It's usually minor stuff like some selective color (pull some yellow out of the reds).
Andrew Rodney
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Sun, Jul 22, 2001, 3:24 PM
RE: Re: [colortheory] reasons why profiles suck so bad
>Chris, I can think of a bunch reasons why profiles
>would "suck so often..." , quite aside from device
>inconsistency. Mainly, the problem I see is in the
>handling of out-of-gamut colors. The current crop
>of cheap flatbed-scanner-based printer-profiling
>systems are prone to yielding sucky results. I've
>been trying for a year or more to produce a profile
>that won't turn my deep blues into deep cyan.You get what you pay for. When they work, they work pretty well considering the cost. But they seem to work less often than not, and when they don't work the results are pretty bad.
However, this is not an example of why profiles suck, nor is it an example of what Dan said:
>profiles generated by a series of machine measurements suck
1. Just because the profiles you generated with one cheap package doesn't mean all profiles suck. It means the profile you made with this one package sucks.
2. The profile you made wasn't made using a measurement device. It was made using a scanner. You introduce scanner metamerism with this method of generating profiles. This isn't something that happens with measurements based on device designed for making Lab based measurements.
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Sun, Jul 22, 2001, 3:43 PM
RE: Re: [colortheory] reasons why profiles suck so bad
Ron Bean writes:>What does "generated by a series of machine measurements" mean to you?
>(Or more to the point, what does it mean to Dan?)It means to me a profile made from a measurement device designed for making color measurements. A minimum would be a colorimeter. Ideally it would be a spectrophotometer. (i.e. a scanner does not qualify because of the rather large likelihood of scanner metamerism that can occur.)
>Weren't you telling me a while back that profiles have to be
>tweaked by hand? (Or maybe that was Chris...) In any case, how
>about a little less heat and a little more light...If the output device has good consistent behavior, and you don't have any errors measuring the target used for making a profile, it will work very well. Saying profiles generated by a series of machine measurements suck is a misrepresenttaion. They don't suck. I would classify them as usually being good, sometimes just OK. Whether they are good enough is a different matter, and for the best possible profile you will need to edit them. But that doesn't mean unedited, machine generated only profiles SUCK. That's completely false and I challenge Dan to prove otherwise.
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Sun, Jul 22, 2001, 6:48 PM
RE: Re: [colortheory] reasons why profiles suck so bad
>I'm not sure I got what I paid for. I've yet to
>get one useable profile from the package (the
>original "Matchlock,", which is now Profiler RGB.)Well you have to be a consumer, just like with everything else you buy. Did you call their tech support? What did they recommend? If it was unsatisfactory did you ask for a refund within a reasonable period of time (such as 30 days from the date of purchase)? Did you ask about refund policy before buying the product? These are not color mangement issues, these are consumer issues. You still have to look out for yourself.
>I believe in your post (a couple of posts back) you implied
>that inconsistency (of the device being profiled) was the main
>reason for bad profiles.That and operator error (measuring the wrong strip for example) are the main reasons for bad profiles.
> I simply wanted to point out that
>there are other, quite obvious causes of bad profiles.Scanner based methods of making profiles is low end. You get what you get. If it works, you saved a few thousand dollars compared to the types of products that we know work when the device is consistent. If it doesn't work, oh well. It's like gambling. The cheap products are a crap shoot. If you look at the numbers of profiles made in the world, scanner metamerism doesn't occur very often because most profiles aren't made with scanners.
>I did work briefly with a borrowed DTP-41, and an "eval" copy
>of Profiler Pro, and managed to make one or two useable profiles.
>I say useable, though not great, and not better than the canned
>profiles I get with my printer.I haven't used Profiler Pro so I don't know how good the profiles are. However I have used a DTP41 and several other instruments with other products including ProfileMaker Pro from Gretag Macbeth and I have yet to make at least a good if not excellent profile unless I did something wrong or the device wasn't behaving consistently. So that tells me that this issue has nothing to do with profiles being based on measurements like Dan was implying, but has more to do with the package that makes the profiles.
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932
From: Dan Margulis, INTERNET:76270.1033@compuserve.com
Date: Mon, Jul 23, 2001, 1:03 AM
RE: [colortheory] reasons why profiles suck so bad
Chris Murphy writes,>>The only reason I'm aware of as to why a profile would "suck so often and so badly" is because the device behavior is inconsistent, and therefore a profile can't be made quickly enough to adequately describe a device's behavior. Maybe you'd like to be a little more clear on what you're trying to say?>>
There are a number of reasons why machine-generated profiles SSOASB, but the only one I was referring to was the point brought up by David, that in color matching, accuracy in the A channel is far more important than in the B. Many of your calibrationist friends do not realize this, and measure how close their colors are using an equation that rates the two as being equal.
Dan Margulis
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Mon, Jul 23, 2001, 10:30 AM
RE: Re: [colortheory] reasons why profiles suck so bad
Dan writes:>There are a number of reasons why machine-generated profiles SSOASB, but
>the only one I was referring to was the point brought up by David, that in
>color matching, accuracy in the A channel is far more important than in the
>B. Many of your calibrationist friends do not realize this, and measure how
>close their colors are using an equation that rates the two as being equal.It doesn't matter which reason you're referring to. The conclusion that machine-generated profiles suck is not only false it's ridiculous to even say it. If you get a profile that sucks, I can promise you that it being based only on measurement data will not be the reason why it sucks unless there is something wrong with the measurement device.
It's fair to say that profiles based on measurement data only, don't do the best job. The best job possible would be a profile that has been properly made in the first place, but then has also had a series of edits applied based on visual evaluation.
You are stipulating that they suck because they are machine-generated. This is so ludicrous it's laughable. There are a number of reasons why some (perhaps even many) profiles suck, but neither measurement data only profiles, nor weighting the a* and b* axes the same would be reasons why. Those would be reasons why they aren't excellent, and only "good". They cannot make such a huge difference to result in something that "sucks."
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932
From: INTERNET:geoff_shearer@yahoo.com, INTERNET:geoff_shearer@yahoo.com
Date: Mon, Jul 23, 2001, 11:52 AM
RE: [colortheory] Re: reasons why profiles suck so bad
--- In colortheory@y..., Chris Murphy wrote:
> >I'm not sure I got what I paid for. I've yet to
> >get one useable profile from the package (the
> >original "Matchlock,", which is now Profiler RGB.)I have recently used Profiler RGB and Doctor Pro (which in my opinion is essential if you want to use Profiler RGB) to profile a Fuji PG4000 and a Gretag Netprinter that we use in-house. I resisted profile-based color management for a number of years, and I still have many unresolved issues with the technology both in theory and in practice. I have managed to come up with profiles that provide a very good output match between my calibrated screen and these two devices. It took weeks of work in my spare time, probably a dozen emails back and forth to ColorVision, and a lot of trial and error (mostly error)on my part. There have been a number of comments about these "low-end" packages on this thread, but I think that this particular package can create profiles fairly well. Whether it is worth the time to do so (in my case, probably 20-30 hours) is another issue. Certainly this "cheap" package was not very cost-effective for me. However, I hope that the knowledge and experience I gained may prove helpful in some future effort.
Geoff Shearer
Digital Photographer
From: Andrew Rodney, INTERNET:andrew@digitaldog.net
Date: Mon, Jul 23, 2001, 12:30 PM
RE: Re: [colortheory] Re: reasons why profiles suck so bad
on 7/23/01 9:19 AM, geoff_shearer@yahoo.com at geoff_shearer@yahoo.com wrote:> It took weeks of work in my spare time, probably a dozen
> emails back and forth to ColorVision, and a lot of trial and error
> (mostly error)on my part. There have been a number of comments about
> these "low-end" packages on this thread, but I think that this
> particular package can create profiles fairly well.ProfilerPro (which unlike the non pro version that uses a scanner can use a true Spectrophotometer) can create a good profile. But yes, it1s a hassle as there is no direct UI for taking measurement data. I know Color Vision is working on this. So it1s a bit convoluted to measure spectral data into one application (with a DTP-41 it1s tool crib which is pretty brain dead), then transfer the data to a text file and import into Photoshop to use ProfilerPro. If you measure anything incorrectly with a device like a Spectrocam, your profile sucks (and this is an issue with improper measure technique so it ends up being user error).
Andrew Rodney
From: "Ron Bean", INTERNET:rbean@execpc.com
Date: Mon, Jul 23, 2001, 1:28 PM
RE: Re: [colortheory] reasons why profiles suck so bad
Andrew Rodney writes:>What do you mean, "Machine-Generated profiles"?
Just to clarify, I mean a profile generated entirely by machine. If you've tweaked it by hand, then it's *partially* machine-generated. Obviously this requires more effort/skill/experience than just reading the numbers from the machine.
I don't know if that's what Dan meant, though. I just read it that way because it relates to a question I asked earlier. Chris's statement below is consistent with what he said back then:
>It's fair to say that profiles based on measurement data only, don't do
>the best job. The best job possible would be a profile that has been
>properly made in the first place, but then has also had a series of edits
>applied based on visual evaluation.
From: Chris Murphy, INTERNET:lists@colorremedies.com
Date: Mon, Jul 23, 2001, 4:40 PM
RE: RE: [colortheory] reasons why profiles suck so bad
>Would you expect a profile tweak if the project is art prints that are then
>going to be printed offset in limited editions with utmost critical review
>or, further, would you expect that each, individual image will be tweaked
>(either before or after the conversion to CMYK)? I don't expect any sucking
>sounds, but I do expect some little tweak, tweak sounds.In a case like this I expect there will be quite a bit of proofing in order to ensure the intended reproduction occurs. The best profile editing for press output occurs over time so that it's possible to establish trends. All you can really expect from profile editing is to fix problematic trends, like reds always have too much yellow or highlights have too much magenta. You can fix these kinds of things with profile editing.
I wouldn't say, in an extremely critical environment, that you're going to have a nice great RGB image on screen, preview as CMYK on screen, and go direct to press and get exactly what you want. You'd going to need to generate proofs, and you're going to need to fine tune the files using CMYK color correction. You would have to do this with any workflow - BUT at least when using accurate profiles, the bulk of the initial work can be done trusting your monitor without having to pull some kind of proof, whereas in a conventional workflow you would.
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932
From: Axel Rose, INTERNET:rose@sj.com
Date: Tue, Jul 24, 2001, 6:57 AM
RE: Re: [colortheory] reasons why profiles suck so bad
Hello all,I'm observing profiles where the *a/*b values are out of there specified range i.e. smaller than -127 and larger than 128. Furthermore there are profiles where *a/*b are interpreted reversely.
This leads me to the conlcusion that implementors rather than machines are often not so clear about *the* basis of ICC color management. Obviously they didn't check the results of their conversion algorithms.
just my two cents :)
Axel
-- ----------------------------------------------------------------------
Axel Rose, Springer & Jacoby Digital GmbH & Co. KG, mailto:rose@sj.com
pub PGP key 1024/A21CB825 E0E4 BC69 E001 96E9 2EFD 86CA 9CA1 AAC5
"Was man nicht weis, das eben braucht man.
Und was man weis, kann man nicht brauchen."
-------------------------------------------------------------------------------
From: "Bruce J. Lindbloom", INTERNET:blind@picto.com
Date: Tue, Jul 24, 2001, 12:05 PM
RE: Re: [colortheory] reasons why profiles suck so bad
Axel Rose wrote:
> I'm observing profiles where the *a/*b values are out of there
> specified range i.e. smaller than -127 and larger than 128.First of all, I think you mean -128 to +127.
Secondly, if you are talking about ICC profiles, the a* and b* ranges are:
-128 to +127 ( 8-bit encoding) -128 to +127.996 (16-bit encoding)
There is no possible way in ICC format to represent a* or b* values that are outside these ranges, due to the encoding methods dictated by that specification. It's impossible.
Thirdly, understand that the "specified range" limits on a* and b* shown above are imposed by the encoding method used by a particular data format, such as ICC format or Lab TIFF format. These limits are not imposed by the Lab color system itself. There are real colors whose a* and b* components are outside the above ranges. Granted, there are not many, but there are some. I have written a one-page tutorial on this topic, which may be found on page 24 of http://www.picto.com/ColorSynergy45Guide.pdf for those interested in exploring further.
Axel Rose continued:
> Furthermore there are profiles where *a/*b are interpreted reversely.I would be very interested in seeing the profiles in question. What software created them? There are no mainstream profiling packages that I know of that make any such blatant errors. Please e-mail me some examples off-list.
Bruce J. Lindbloom, Pictographics Intl. Corp.
From: "Dave King", INTERNET:kingphoto@mindspring.com
Date: Wed, Jul 25, 2001, 2:58 PM
RE: Re: [colortheory] Re: reasons why profiles suck so bad
> I have managed to come up with profiles that provide a
> very good output match between my calibrated screen and these two
> devices. It took weeks of work in my spare time, probably a dozen
> emails back and forth to ColorVision, and a lot of trial and error
> (mostly error)on my part. There have been a number of comments about
> these "low-end" packages on this thread, but I think that this
> particular package can create profiles fairly well. Whether it is
> worth the time to do so (in my case, probably 20-30 hours) is another
> issue. Certainly this "cheap" package was not very cost-effective
> for me. However, I hope that the knowledge and experience I gained
> may prove helpful in some future effort.Some of the problems people have with Profiler RGB and other scanner packages are apparently due to scan driver interface issues. When I was using Profiler RGB, if the scan driver wasn't set correctly (and correct setting wasn't that apparent in my case), the results were pretty much like the usual complaints -- cyan blues, posterized shadows etc. Unusable unless psychedelics are your thing. When set correctly however the profiles were pretty good right out of the box, not as good as Epson oem dye printer profiles, but useable, and making possible 3rd party pigments for a pretty low investment. But it is telling how much effort even expert users put into modifying scanner profiles to bring them to the level of excellence. Not perhaps for the faint of heart, or color sensitive amateur.
Upgrading to Profiler Pro and a DTP-41 was an eye opener, as the profiles are excellent right out of the box, though some profiles will need a minor tweak still. I tried a hand held measurement device initially (CM2C ColorMouse) and it drove me crazy with user errors on top of taking forever, so I got a DTP-41. The DTP-41 is considerably more accurate than the ColorMouse (even when that device is used correctly), and building a profile takes about 10 minutes. It's a doddle, and no errors.
It seems to me that no profile can completely close the gap between phosphor and ink, but I was initially astonished at how close monitor and page appear with accurate calibration. And most importantly to me, I feel a combination of correction by the numbers *and* direct viewing of color on a calibrated system gives the potential for a better correction than would either method used alone. This may have to do with my inexperience level, but it seems to me there's no substitute to viewing color interactions (global color contrast esp) for the more critical corrections.
Dave King
From: Axel Rose, INTERNET:rose@sj.com
Date: Wed, Jul 25, 2001, 5:27 PM
RE: Re: [colortheory] reasons why profiles suck so bad
Hello Mr. Lindbloom,thank you for your reply.
I'm only a casual user of ICC based color management and have to admit that my remarks about malformed values were a bit naively. They were driven by reports from other users. I'll try to clarify this later.You are abolutely right about the value range of *a/*b within an ICC profile. The coding scheme makes excessive values impossible. (not so for 16bit L by the way)
I suppose though that invalid measurements or software algorithms can easily lead to out of range values within intermediate files.
I was astonished to read in the excellent ColorSynergy user guide that only 35% of all the possible L*a*b values represent real colors. I'd like to know in how far this affects the accuracy of profiles which is the concern of this thread.
My understanding furthermore is that there are two questions -- firstly is it easy to produce erroneous profiles (no doubt I think) and secondly whether current ICC profiles contain inherent inaccuracies.
Best regards
Axel Rose
Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.