Re: RelCol Mapping of Middle Gray
Re: RelCol Mapping of Middle Gray
- Subject: Re: RelCol Mapping of Middle Gray
- From: Steve Upton <email@hidden>
- Date: Mon, 7 Mar 2005 19:20:47 -0800
At 4:19 PM -0800 3/7/05, Michael Fox News Accout wrote:
>I'm curious about the RelCol intent mapping of the gray axis and, specifically, middle gray. To start with, we know several things that are immutable facts:
>
>RGB = 128,128,128 = middle gray
>Reflective Density = 0.78,0.78,0.78 = middle gray
>In LAB, a*b* = 0,0 is neutral. It seems that the L value for middle gray is dependent on the color space.
>
Actually I would think you have these things reversed
RGB=128,128,128 is probably NOT middle gray. If we are talking about a working space then the gamma of the space will determine the lightness of the 128,128,128 color. In 2.2 gamma spaces (sRGB, Adobe RGB) the L value is 54. In 1.8 spaces (ColorMatch) the L value is 61. You need a gamma of something like 2.43 to get an L value if 50
I'm not an authority but doesn't reflective density depend on paper white and 100% solid? It would seem like that value would vary as well
In Lab, 50,0,0 IS middle gray by definition. So this is one of the values that would NOT change.
>Given those things, if I profile my printer and I send it a file that has RGB=128,128,128, I should be able to measure the result with a reflective densitometer and get 0.78,0.78,0.78. Alternatively, if I measure the result with a spectrophotometer, I should get a*b* values of 0,0.
>
well, no. But yo also need to tell us what 128,128,128 you are sending the printer.. working space or raw printer RGB? Unless your working space has a gamma of 2.43 then you won't get middle gray. Also, black point compensation can affect this.
>But, I don't. Generally, what I get is a*b* values that match the paper white value. For example, if paper white is something like LAB = 96, 0.4, -1.7, then I'll get a middle gray that is fairly close to middle gray density of 0.78 but has a color value in a*b* of 0.4, -1.7 (+/- 0.1).
>
is was good of you to observe that it is similar to paper white. Many would have missed that.
>I say generally because I observe this behavior with Monaco Profiler Platinum on different papers and with different printers. I remember the last profile I made with Gretag ProfileMaker was off in color and density but I don't recall if it matched up exactly with the color of paper white like the Monaco software seems to be doing.
>
>Now, I can understand that as the L value approaches paper white, the colors will be more and more influenced by paper white. But at middle gray, there should be enough ink to produce a color that is neutral. In fact, I can edit the profile to make it that way. But I don't think I should have to do this.
>
no, but as Bruce suggested, you normally wouldn't want to do it either. Once your eye adapts to paper white you want the grays to be built using mixtures of black and paper white, otherwise you get shifts up the gray ramp.
>I further understand that Monaco and Gretag have options for how to handle the gray axis in the Perceptual intent. But I'm talking about the Relative Colorimetric intent. I would have thought that RelCol would map a neutral in the image file to a true neutral in the print. Perhaps I'm wrong.
this was something i was confused about at first when I started graphing colors & color vectors. The important thing about relative colorimetric is that it maps color to the paper RELATIVE to the paper's white point. So, in fact, it is changing all colors in the transformation - light ones more than dark ones. Abs col is the one that should give you colors closer to what you are expecting... it can have its issues as well though...
At 8:47 PM -0500 3/7/05, Ken Fleisher wrote:
>Actually, it sounds like you are getting the correct values. Remember that ICC L*a*b* is NOT the same as CIE L*a*b*.
>
>In ICC L*a*b*, the white point of the media is used instead of the white point of D50 (it's assuming that you are chromatically adapted to the white point of the media). Therefore values of L*=100, a*=0, b*=0 represent, by definition, the white point of the media, not the light source.
actually, I don't think that's true. The white point of Lab is D50 but D50 is NOT the paper white. The paper white will be a number relative to D50 (as if measured/viewed under D50) but will probably be something like 96,1,3.
> If you want to make a direct comparison using your spectrophotometer and ICC L*a*b* values, then you must also factor in the white point of your media when calculating L*a*b* (which your spectrophotometer is NOT doin--if that is where you are getting the values from). The fact that you are getting the same a*b* values as your white point is therefore correct.
Ah! I see what you are talking about... OK then yes... sort of. IF you want to compare measurements from two different paper types to see if they match (not side-by-side but after the viewer has adapted to each of them) then yes, you can pull the paper white out of the measurements using XYZ scaling
>To calculate into ICC L*a*b*, you can take your XYZ values from your spectrophotometer (D50 and two degree observer) and convert to L*a*b* using the media white point (instead of D50). This should then result in a*b* values of 0 for neutral colors and match the values that the profile is trying to make.
I guess you could do that too but is that the typical way of doing the conversion? I implemented that a while back... I'll have to go check. I thought you just normalized Y....
Regards,
Steve
________________________________________________________________________
o Steve Upton CHROMiX www.chromix.com
o (hueman) 866.CHROMiX
o email@hidden 206.985.6837
o ColorThink ColorValet and now... ColorForums.com
________________________________________________________________________
--
_______________________________________________
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