Victor,
working space and monitor space are defined for white point D65. Prints are optimized for D50 in a way that the appearance is VERY SIMILAR to the monitor.
I don´t think that you can directly shift the color temperature of neutral gray in an image by PhS,
but I am not a PhS expert.
The shift to D50 (red-ish look) or to D93 (blue-ish look) is done by a matrix multiplication RGB-new=M*RGB-old.
The accurate algorithm is available, but for simple tests this will be be sufficient: (1) for D65–D50: multiply R by 1.1765, G by 0.9765 , B by 0.7218 (2) for D65–D93: multiply R by 0.8986, G by 0.9938 , B by 1.3602 By the way, this will cause clipping problems for Red (1) or Blue (2) . Green is not much modified.
Perhaps it helps to adjust an image towards the one direction or the other.
Best regards –Gernot Hoffmann
Victor,
forgot to add that this has to be done in the linear RGB working space. R=R´^gamma. Apply modification. R´= R^(1/gamma).
G.H.
Oke understand it Gernot,
One thing though; I understood that PhotoShop was using sRGB with D50, instead of the standard definition of sRGB at D65. You are now saying it is sRGB at D65 for PhotoShop, are you sure?
See also:
<
http://www.brucelindbloom.com/WorkingSpaceInfo.html> Section on ‘Adapted primaries’
It is all quite confusing to me (not being an expert);-) I am trying to do some color optimization for a standard scale made by IFRAO see: <
http://www.iol.ie/~geniet/eng/IFRAOcoloropt.htm>
and for this I try to get a good idea what happens. Any input is VERY welcome.
All the best,
Victor
No, Photoshop uses sRGB exactly as it is defined in the IEC standard.
You have many mistakes in your web page.
1) sRGB always uses a D65 white point (as given in the IEC standard), and Photoshop uses sRGB exactly as given by the standard.
2) You can do gamma adjustments in Photoshop (there is no need for ImageReady in your steps)
3) Photoshop does not contain a capital S, ever
4) You don’t need Brightness/Contrast at all — just Levels
5) Photoshop 4 didn’t have ANY color calibration, so you couldn’t reproduce sRGB in it.
6) The TRC/gamma functions used in sRGB and most other standard color spaces are not pure power functions – they include some slope limiting to improve matching to human vision and provide invertibility for the TRC function (otherwise the inifinite slope of the power function messes things up).
You really should compare the accuracy of your method to other calibration methods (typically using many more color samples, such as an IT8.7/2 chart or a MacBetch ColorChecker DC).
Photoshop does not contain a capital S, ever
Except when you type PHOTOSHOP. Sorry, couldn’t resist.
Victor,
formulas and a diagram for the sRGB Tone Reproduction Curve are e.g. here: <
http://www.fho-emden.de/~hoffmann/gamuts08072002.pdf> (150kB)
My suggestion for WP shifts didn´t contain anything about adaptation. This test is valid:
Working space sRGB (with D65 according to standard) .
Make background in image neutral gray: R=G=B=118 (or so). Measure color temperature by instrument: nearly 6500K.
Convert image to D50.
Measure color temperature by instrument: nearly 5000K.
And so on. The test requires an accurately calibrated monitor. A set of test images D50/D65/D93 is here:
<
http://www.fho-emden.de/~hoffmann/caltutor270900.pdf> (550kB)
By the way: I´m not trying to automate image processing. The only exception are WP corrections if dominating neutral gray areas are available. This is very helpful for digital photos.
Best regards –Gernot Hoffmann
Hello Chris,
Thanks for your feedback. This is very valuable to me!
Can you be sure about the D65 (e.g. I know Paint Shop Pro is using indeed D65)? To be honest I had a big discussion about this before I changed it to D50 for Photoshop (see also the earlier URL given, according to this source Photoshop’s sRGB uses the same definition as ICC for the white reference: D50).
Perhaps I should directly ask Photoshop! But can you provide some URL’s where it is stated that they use D65 (I tried to look in the help files but found nothing specific).
Thanks for the feedback what can be done with Photoshop. I was relaying on someone else here. I now downloaded it myself so that I can check things myself.
I understand from several sources that most RGB workspaces are pure power functions. One exception which is the sRGB, which has a linear part near the black region (so for low values it is not a power functions). I got this from URL’s like:
<
http://www.poynton.com/papers/IST_SPIE_9801/index.html>
I agree, I will to check this all with better charts, I have pictures with the Macbeth ColorChecker and IFRAO standard scale. But first I need to get my basics correct (likr the D? and standard observation angles, so measuring with colorimeter is at this moment being done).
I have changed all your other remarks in my text.
Thanks again and all the best,
Victor
Hello Chris,
Can you tell me where I can adjust the gamma in Photoshop? I can’t find it? I still want to stay in the sRGB environment though. In ImageReady I can do it, but see nothing like that in Photoshop.
All the best,
Victor
Hello Chris and Gernot,
Here is some sRGB information: <
http://www.srgb.com/> and: <
http://www.srgb.com/c55.pdf>
This last link is clearly stating that D50 should be used for sRGB and I understood Adobe is doing that also in the way it is stated in that document. Do you have contacts with Adobe to verify this? I have no support contract or something like that.
All the best,
Victor
Victor,
the second link is well-known. Thanks for both.
sRGB is primarily a working space, well defined by three primaries, D65 white point and a Tone Reproduction Curve (TRC) which is near to Gamma=2.2 (less 1% deviation).
The authors are talking ADDITIONALLY about the question how to transform sRGB data into a so- called Profile Connection Space (PCS), which can be used as input for multi-dimensional look-up tables with CMYK outputs for printing.
Graphics art is traditionally viewed "under" D50 light. The aim is: the printed paper under D50 should
look equal to the monitor view. The monitor can reproduce sRGB more or less directly, if calibrated for D65.The (simplified) conversion is done by the Bradford matrix. There is an essential adaptation problem: if monitor and print are viewed one after the other, then eye+brain are adapted alternating to D65 and D50 and the perception of the SAME physical color would be different. The balance for equal appearance is (hopefully) done by Bradford. sRGB itself is defined for D65 white point.
The appreciated authors shouldn´t assign D50 to "sRGB for printing". These transformations be-
long to the PCS, IMO, and to the definition of the two viewing conditions.
There is hardly anywhere an algorithmical and understandable explanation. The ICC specification is still rather confusing (after reading the important parts twenty times).
Best regards –Gernot Hoffmann
Hello Gernot,
Thanks for your explanation, I hope I understand it. So now in my words;-). I understand sRGB is defined as being in the D65 space as according to the standard. I also understand the e.g. a programa like Paint Shop Pro is also using this. From one person (Bruce Lindbloom, who is quite knowlegdable on this subject) I heard that Photoshop does not use the default white point for sRGB but uses D50 (or perhaps I misunderstand his pages: <
http://www.brucelindbloom.com/index.html?WorkingSpaceInfo.ht ml> section ‘Adapted Primaries’).
I understand from you and Chris that this is not true, but that Photoshop is using for sRGB, as defined in the standard, D65.
So this is where I am stuck. I hope that Photoshop can provide the final answer on this (I have no support contract, so I am not able to ask this officially). Are you willing to do this?
In the meantime, I have changed my web-page in such a way , that I don’t link Photoshop anymore to the D50 (except in a note with a question mark). I have now two columns one for sRGB/D65 (standard) and one for sRGB/D50 (non standard). As soon as I know what precisely Photoshop uses, I will update the table.
Thanks again for your valuable input!
All the best,
Victor
Victor – I don’t have URLs, I’m looking at the source code and the sRGB profile. (there is a reason my name appears in the splash screen)
No, most profiles do NOT use pure power functions. In fact, I don’t know of anyone other than yourself (and one net troll) who makes that mistake.
The Gamma adjustment is the middle slider in the levels dialog, same as in ImageReady.
I don’t know how you’re getting D50 for sRGB – it is very clearly using D65, always. I’ll have to check with Bruce and see how he made a mistake thinking that Photoshop uses something other than the standard sRGB profile.
OK, Bruce has it correct. But you are misreading what he wrote.
In ICC profiles, all the primaries are adapted to the PCS whitepoint – which is defined as D50. The sRGB whitepoint (in a separate tag) is still D65.
Hello Chris,
I am glad you can confirm that the sRGB in Photoshop is indeed using D65. It was what I always thought, but could not get the correct arguments in place towards Bruce (even though it is in his first table!). I think we had a basic misunderstandig, due to my inexperience perhaps.
Thanks for confirming Photoshop’s implementaion!
Now that I am certain about the workspace and how to convert them using colorimeter measurements, I will improve my optimization method. Including another gamma function is not problem, and hopefully it will make the optimization process even more stable.
The gamma option in Photoshop is very essential, so I am glad it is there (will check this night). Now the use of Photoshop becomes much more fluent.
I hope that my methodology of course works for this simple standard scale (i know about other scales, but it seems that the IFRAO scale is also widely used). I had my personal doubts on its effeciveness, two doubts I had: a) no proper reference color definitions (so only RGB values of Photshop4 were existing) and b) NO procedure how to use it for optimization!!!
I am now trying to solve these points. And your feedback is helping me greatly.
I am just a layman wanting to be sure that my time in using the IFRAO card at rock/megalithicart is not wasted!
All the best,
Victor
Hello Chris,
But thanks to you, I found the Gamma function.
I re-checked the help files, the word ‘Gamma’ is not used in the Levels dialog help files. It is differently describe as: "You can use the middle Input slider to change the intensity values of the middle range of gray tones without dramatically altering the highlights and shadows." That made me miss the relation with Gamma, sorry (Photoshop was up to now not my default program).
Another thing, I can do different functions for gamma’s in my optimization program, no problem! But does Photoshop support different gamma functions (so with linear parts), looking at the supported gamma in the Levels it seems to be the pure power functions.
To be realisitic my optimization program uses the same functions as available under Photoshop/PSP.
All the best,
Victor
I would like to add a question:
According to my test program IccInspect by LittleCMS the working space Adobe RGB (98) uses a pure power function with exponent 2.2 without a linear slope. Is this a misinterpretation by IccInspect ?
The tone reproduction function can be defined by
a) 3 tables
b) 3 parametric functions of different complexity.
Best regards –Gernot Hoffmann
The gamma adjustment in Levels is a predefined power function with a linear slope, plus a small spline segment to smooth the meeting of the two functions.
Sorry that the documentation left out the word gamma. We’ll have to work on that.
Gernot – the ICC profile spec’s it as a power function, but all CMMs that I know of add the linear slope.
Chris,
"linear slope for Adobe RGB(98)"
This not according to the standards. The CMS has to use the features as defined in the profile.
Best regards –Gernot Hoffmann
Hello Chris,
Can you provide me the forumula of the Gamma function which is defined in the Photoshop Levels. I really would like to see this, because I have not seen this earlier.
It is also important for people to know, I would say, that PhotoShop is not using the standard Gamma function (I don’t want to confuse this with the specific Gamma functions defined for workspaces).
Checking the numbers, I know for sure PSP is doing a real Gamma (at least the rounded values to integers are one to one), I will check this night what the numerics in Photoshop (I did not see much descrepancy, so I thought it was according to a pure Gamma function).
But if you can provide the function that would help me (also for my optimization program).
All the best,
Victor
Victor — sorry, but I’m not allowed to give you the code or exact algorithm.
Again, most applications do NOT use a pure power function (that’s a bad idea), they use a power function with slope limiting.
And no, the war with TImo started with his many lies and his campaign to spread misinformation. And that started almost a year before I joined Adobe, on usenet (where Timo had other wars with experts in Photography, Television/Video, Human Vision, Color Science, etc.).
Chris,
if Adobe RGB(98) uses a linear slope though the profile says it´s a pure power function, then other color managed programs can´t reproduce images, which were edited in Adobe RGB, correctly.
Best regards –Gernot Hoffmann
Victor,
as far as I understand the CM thing, you have to profile your image. If you want it to look like it might look at a different light temperature, you have to assign a profile. If you want to change the white reference of the picture itself, you have to convert it to a profile. You can create in Photoshop your own profiless or modify an existing profile e.g. by changing its WP.
Dennis.
Hello Chris,
Perhaps you don’t like me quoting this link:
<
http://www.aim-dtp.net/aim/evaluation/gamma_error/main.htm#l evels> but the problem is that indeed I experience the same problems in Photoshop7, and because you don’t provide me with the mathematical function, I can’t align my optimization program with Photoshop, which I find a pity (because it is the most used program in the IFRAO environment).
It seems this behavoir is most effecting high Gamma’s (not so when using gammas lower then 0.5).
I really hope that you can help in some way. Are there alternative commands I can use and get the real power function Gamma (if you don’t publisize your functions)?
All the best,
Victor
Victor – You’re right: I don’t like people quoting misinformation sites. And that one is a misinformation site.
Gernot – it doesn’t say it’s a pure power function. It says it’s a gamma function characterized by a single number, that all available CMMs interpret as a power function with slope limiting.
Hello Chris,
But can you help me then with my problem stated (irregardless of the quoted site, which happened to mirror my own findings on this gamma issue)?
Is there somewhere in Photoshop 7 a pure power function available or can you make Photoshop gamma function available?
All the best,
Victor
Chris,
the profile says, theTone Reproduction Curve for Adobe RGB(98) is a pure power function with exponent 2.2. This according to IccInspect by LittleCMS. Without additional information this cannot be interpreted as a power function plus linear slope.
ICC specs distinguish clearly between TRCs by tables and by parametric functions with different degrees of complexity.
Adobe RGB(98) belongs IMO to the simplest cases: C=C’^gamma, or the inverse function.
If I am wrong then it should be proved. If IccInspect is wrong then it should be proved. The odd TRCs as published by Timo Autiokari are showing that PhS seems to apply a proprietory interpretation which is not complient with ICC specs.
I don´t worry about, but the issue shouldn´t be blurred intentionally.
IF THE PROFIL DEFINES A PURE POWER FUNCTION THEN IT HAS TO BE HANDLED BY PHS EXACTLY LIKE THIS. OTHERWISE DEFINE A NEW WORKING SPACE.
Best regards –Gernot Hoffmann
Gernot – apparently you’re missing the point of practice versus theory.
In theory it’s a power function, in practice it’s a power function plus slope limiting (because most people know better than to use just a power function).
Chris,
so you are saying that e.g. Onyx or Best RIPs apply a linear slope if Adobe RGB(98) is chosen as source space for RGB images, though the profile contains a pure power function ?
The difference is of course relevant for the internal conversion to CIELab and would change dark colors considerably.
Best regards –Gernot Hoffmann
Gernot – yes. The profile specifies a single number, not a pure power function.
