Physics Talk Physics Talk    FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in

Education Forums

[Q] Convert XYZ to RGB

Physics Talk Forum Index -> Color and Light

View previous topic :: View next topic
Author	Message
Eric Guest	Posted: Fri Nov 16, 2007 3:38 pm    Post subject: [Q] Convert XYZ to RGB My source data is the Munsell real.dat file found at: http://www.cis.rit.edu/mcsl/online/munsell.php and I need to convert this data into the sRGB color space. I am pretty sure I know how to accomplish this, but was hoping for some confirmation that my method is accurate. The method I am using is based on the sample code found at: http://casoilresource.lawr.ucdavis.edu/drupal/node/201 The primary question I have is how to correctly apply a chromatic adaptation transform to the XYZ color to convert it from its own reference white to the reference white of the RGB system - assuming I need to. The Munsell data was apparently calculated using illuminant C. What is the reference white of the sRGB system? Is it D65? If so, I believe I would want to apply the following matrix: 0.990490 -0.012269 -0.003515 -0.007115 1.015427 0.006679 -0.011434 -0.002878 0.919313 This is the Bradford matrix found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html The sample code I pointed to above does not appear to apply this matrix and I believe that would be considered a bug which I would want to fix in my own code. The Munsell data also mentions that it was calculated using the 'CIE 1931 2 degree observer'. So, my other question is how I would take this into account in my conversion - assuming I need to...? The general procedure for converting between XYZ to RGB found at http://www.brucelindbloom.com/index.html?ColorCalcHelp.html doesn't mention observers, but I wanted to make sure that it would not affect the calculations and perhaps to understand what this means a little better. Any thoughts would be appreciated. Thank you. p.s. The munsell data actually starts in xyY, but the conversion to xyY to XYZ appears to be straightforward using the formula found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html . Furthermore, the munsell data appears to be in the range of 0.0 - 100.0 which is why I believe the sample code above is doing the division by 100.
Back to top
	Ads
Advertising Sponsor
Eric Guest	Posted: Sat Nov 17, 2007 5:48 pm    Post subject: Re: Convert XYZ to RGB On Nov 17, 11:00 am, Roger Breton <gr...@videotron.ca> wrote: Quote: Eric, You may want to take a look at BabelColor's CT&A atwww.babelcolor.comfor pointers on how to convert from XYZ to Munsell. Thank you. However, if you can answer any of the questions directly, I would appreciate it.
Back to top
	Ads
Advertising Sponsor
Roger Breton Guest	Posted: Sat Nov 17, 2007 10:00 pm    Post subject: Re: [Q] Convert XYZ to RGB Eric, You may want to take a look at BabelColor's CT&A at www.babelcolor.com for pointers on how to convert from XYZ to Munsell. Roger Breton
Back to top
	Ads
Advertising Sponsor
Gernot Hoffmann Guest	Posted: Sun Nov 18, 2007 5:26 pm    Post subject: Re: [Q] Convert XYZ to RGB Eric schrieb: Quote: My source data is the Munsell real.dat file found at: http://www.cis.rit.edu/mcsl/online/munsell.php and I need to convert this data into the sRGB color space. I am pretty sure I know how to accomplish this, but was hoping for some confirmation that my method is accurate. The method I am using is based on the sample code found at: http://casoilresource.lawr.ucdavis.edu/drupal/node/201 The primary question I have is how to correctly apply a chromatic adaptation transform to the XYZ color to convert it from its own reference white to the reference white of the RGB system - assuming I need to. The Munsell data was apparently calculated using illuminant C. What is the reference white of the sRGB system? Is it D65? If so, I believe I would want to apply the following matrix: 0.990490 -0.012269 -0.003515 -0.007115 1.015427 0.006679 -0.011434 -0.002878 0.919313 This is the Bradford matrix found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html The sample code I pointed to above does not appear to apply this matrix and I believe that would be considered a bug which I would want to fix in my own code. The Munsell data also mentions that it was calculated using the 'CIE 1931 2 degree observer'. So, my other question is how I would take this into account in my conversion - assuming I need to...? The general procedure for converting between XYZ to RGB found at http://www.brucelindbloom.com/index.html?ColorCalcHelp.html doesn't mention observers, but I wanted to make sure that it would not affect the calculations and perhaps to understand what this means a little better. Any thoughts would be appreciated. Thank you. p.s. The munsell data actually starts in xyY, but the conversion to xyY to XYZ appears to be straightforward using the formula found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html . Furthermore, the munsell data appears to be in the range of 0.0 - 100.0 which is why I believe the sample code above is doing the division by 100. The linear Bradford CAT for your application is based on the assumptions, that the observer of Munsell samples views these samples under illuminant C and that he(she) is adapted to illuminant C and that he/she views sRGB samples under illuminant D65 (or on a monitor with D65 white point) and that he/she is adapted to D65. Obviously it's very doubtful whether all these assumptions were correct for real world tests. Consider just the fact that many people are viewing their sRGB data on a monitor with white point D50 ... without any correction by color management. D65 cool white appears as D50 warmer white on the monitor, but the CAT doesn't know the monitor white point. Best regards --Gernot Hoffmann
Back to top
	Ads
Advertising Sponsor
Roger Breton Guest	Posted: Mon Nov 19, 2007 1:11 am    Post subject: Re: [Q] Convert XYZ to RGB Quote: Consider just the fact that many people are viewing their sRGB data on a monitor with white point D50 ... without any correction by color management. Like me ;-) Roger Breton
Back to top
	Ads
Advertising Sponsor
Timo Autiokari Guest	Posted: Mon Nov 19, 2007 1:23 am    Post subject: Re: [Q] Convert XYZ to RGB Eric wrote: Quote: The method I am using is based on the sample code found at: http://casoilresource.lawr.ucdavis.edu/drupal/node/201 how to correctly apply a chromatic adaptation transform It looks to me that the sample code does it, even gives two options. Quote: assuming I need to. Yes, it has to be done. Please download the latest ICC spec, it is all explained in there http://www.color.org/ICC1v42_2006-05.pdf (especially in the annex'es at the end). Quote: The Munsell data was apparently calculated using illuminant C. That is what they say on the third row of text. Quote: What is the reference white of the sRGB system? Is it D65? Yes, it is. Quote: The Munsell data also mentions that it was calculated using the 'CIE 1931 2 degree observer'. So, my other question is how I would take this into account in my conversion - assuming I need to...? It means that the spectral data has been converted to trichromatic values by that 1931 model. At this stage (you have only the trichromatic representation) it is just nice-to-know kind of information, and very guessable since it is the de'facto standard. However, the the Std Illumination C is built-in to that data as the whitepoint, this is not the standard practice nowadays. Quote: The general procedure for converting between XYZ to RGB found at http://www.brucelindbloom.com/index.html?ColorCalcHelp.html doesn't mention observers, but I wanted to make sure that it would not affect the calculations Does not, because it has no effect anymore. If you had the spectral data then you'd possibly like to convert them directly to D65 (for absolute colorimetry) or to D50 and from there to D65 by Bradford (for relative colorimetry). Quote: p.s. The munsell data actually starts in xyY, but the conversion to xyY to XYZ appears to be straightforward using the formula found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html. Yes, www.brucelindbloom.com is a very good reference site. The ICC specification is also a good reference. Quote: the munsell data appears to be in the range of 0.0 - 100.0 which is why I believe the sample code above is doing the division by 100. Yes, some want to define the XYZ values in the range of 1 to 100 (or so) and some others some want to use the range 0 to 1 (or so) so there are formulas for both. Using the 0-100 for XYZ makes the XYZ triple to look different than the Yxy triple but adds unnecessary scaling to the formulas. Timo Autiokari
Back to top
	Ads
Advertising Sponsor
Eric417 Guest	Posted: Mon Nov 19, 2007 8:23 am    Post subject: Re: [Q] Convert XYZ to RGB Timo Autiokari <timo.autiokari@aim-dtp.net> wrote: Quote: Eric wrote: The method I am using is based on the sample code found at: http://casoilresource.lawr.ucdavis.edu/drupal/node/201 how to correctly apply a chromatic adaptation transform It looks to me that the sample code does it, even gives two options. It does now. I wrote to the author and he updated the code on the page. Quote: assuming I need to. Yes, it has to be done. Please download the latest ICC spec, it is all explained in there http://www.color.org/ICC1v42_2006-05.pdf (especially in the annex'es at the end). Thanks. I'll check out that reference.
Back to top
	Ads
Advertising Sponsor
Gernot Hoffmann Guest	Posted: Mon Nov 19, 2007 4:15 pm    Post subject: Re: [Q] Convert XYZ to RGB Eric417 schrieb: Quote: I am not sure what you mean by the phrase 'adapted to'...? What does it mean to be adapted to viewing the Munsell samples under different illuminants? Consider just the fact that many people are viewing their sRGB data on a monitor with white point D50 ... without any correction by color management. Wouldn't one consider that if colors were not corrected using the color profile of the device that it would be wrong? Eric, let me start by quoting Mark D.Fairchild, Color Appearance Models. About Munsell, p.98: ' The colorimetric specifications utilize CIE illuminant C and the CIE 1931 Standard Colorimetric Observer (2°). These specifications should be kept in mind when viewing any embodiment of the Munsell system. The perceptual uniformity of the system is only valid under source C, on a uniform middle gray (N5) background, with a suf- ficiently high iluminance level (e.g. greater than 500 lux) . Viewing the samples in the Munsell Book of Color under any other viewing conditions does not represent an em- bodiment of the Munsell system.' OK, these are the test conditions for illuminant C (very difficult IMHO to be established practically). On the other hand, these colors have to be represented in an sRGB system. The most common application is not a lighting by simulated D65 illuminant but a monitor. In order to apply a chromatic adaptation transform, it's assumed that the observer is (on the one hand) adapted to illuminant C and (on the the other hand) to D65. If the monitor is calibrated for D65 and the roomlight is dimmed and the walls are 'white', then we may expect an adaptation to D65. Adapted to a white point means: all grays on a straight line in XYZ from the origin to the specified white point are perceived as neutral grays. Obviously one cannot adapt simultaneously to D50 and D65. A side by side comparison of a monitor image and e.g. a print is IMO not reliable. Considering all these uncertainties, one may ask whether the linearized Bradford CAT is a striking success. IMO it's just a convention (for instance in Adobe programs). Nowhere any proof in printed books 'with Bradford' versus 'with Wrong von Kries'. I would like to add a PDF, showing vector fieldsfor color shifts in the chromaticity diagram xyY, but our server is at present locked. In order to overcome the diffculty of a direct comparison between a D65 monitor for sRGB and D50 proofing light for prints (I'm drifting a little from Munsell to prints), some people got used to calibrate the monitor for D50, with apparently good results. But attention: the CAT doesn't know the monitor's white point (which might be representative for the adaptation state) but assumes still D65. Thus a Bradford CAT is applied, though the two (true) adaptation states are identical or at least near to each other. People are not aware of this because there are so many other effects which affect a pleasant (!) reproduction of color. Best regards --Gernot Hoffmann
Back to top
	Ads
Advertising Sponsor
Eric417 Guest	Posted: Mon Nov 19, 2007 7:31 pm    Post subject: Re: [Q] Convert XYZ to RGB Gernot Hoffmann <hoffmann@fho-emden.de> wrote: Quote: Eric schrieb: My source data is the Munsell real.dat file found at: http://www.cis.rit.edu/mcsl/online/munsell.php and I need to convert this data into the sRGB color space. I am pretty sure I know how to accomplish this, but was hoping for some confirmation that my method is accurate. The method I am using is based on the sample code found at: http://casoilresource.lawr.ucdavis.edu/drupal/node/201 The primary question I have is how to correctly apply a chromatic adaptation transform to the XYZ color to convert it from its own reference white to the reference white of the RGB system - assuming I need to. The Munsell data was apparently calculated using illuminant C. What is the reference white of the sRGB system? Is it D65? If so, I believe I would want to apply the following matrix: 0.990490 -0.012269 -0.003515 -0.007115 1.015427 0.006679 -0.011434 -0.002878 0.919313 This is the Bradford matrix found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html The sample code I pointed to above does not appear to apply this matrix and I believe that would be considered a bug which I would want to fix in my own code. The Munsell data also mentions that it was calculated using the 'CIE 1931 2 degree observer'. So, my other question is how I would take this into account in my conversion - assuming I need to...? The general procedure for converting between XYZ to RGB found at http://www.brucelindbloom.com/index.html?ColorCalcHelp.html doesn't mention observers, but I wanted to make sure that it would not affect the calculations and perhaps to understand what this means a little better. Any thoughts would be appreciated. Thank you. p.s. The munsell data actually starts in xyY, but the conversion to xyY to XYZ appears to be straightforward using the formula found at http://www.brucelindbloom.com/index.html?Eqn_ChromAdapt.html . Furthermore, the munsell data appears to be in the range of 0.0 - 100.0 which is why I believe the sample code above is doing the division by 100. The linear Bradford CAT for your application is based on the assumptions, that the observer of Munsell samples views these samples under illuminant C and that he(she) is adapted to illuminant C and that he/she views sRGB samples under illuminant D65 (or on a monitor with D65 white point) and that he/she is adapted to D65. I am not sure what you mean by the phrase 'adapted to'...? What does it mean to be adapted to viewing the Munsell samples under different illuminants? Quote: Consider just the fact that many people are viewing their sRGB data on a monitor with white point D50 ... without any correction by color management. Wouldn't one consider that if colors were not corrected using the color profile of the device that it would be wrong?
Back to top
	Ads
Advertising Sponsor
Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First

	Physics Talk Forum Index -> Color and Light	All times are GMT
Page 1 of 1

Jump to: Select a forum Physics Talk----------------Physics Talk-- BiophysicsMechanicsColor and LightRelativity  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum

97 Attacks blocked

Powered by phpBB © 2001, 2005 phpBB Group