Re: Colorimetric Accuracy in the Field
Thank you, that's interesting and prompts me to ask: do colorimeters/ spectrophotometers attempt to mimic the human observer? I suppose there can only be a claim for mimicking - for how could it ever be verified? But we use them for measuring and determining a colorimetric match so we have to rely on the claim because besides that there is only subjective observers. Or we rely on both. My question about a measured match between two samples that don't match visually is still important to me and I think the answer puts a lot of stress on how colorimetric accuracy is defined. Henry Graeme wrote: Henry Davis wrote:
Nobody is saying that the illuminant isn't a hugely important factor.
The main reason that the illuminant is a factor is that many/most cameras aren't colorimetric. By that I mean that their spectral sensitivities don't match a human observer. If they did, and their (light level) linearity was known, then (by definition) they'd be a high spatial resolution colorimeter. The practical situation is analogous to the one with cheap display colorimeters: They can be very accurate with a correction matrix if the spectral nature of the samples they are reading is known. With a display device that has colors that are (largely) a liner combination of the three primary spectra, a matrix does nicely. For a camera, it can't be that easy, because real world object colors are composed of many metamers, and the spectrum hitting the camera sensor is that times the illuminant spectrum. So a "calibration" matrix can fudge it for a range of expected object spectra x expected illuminant, but ultimately it's going to work the best over the widest range of conditions if the camera spectral sensitivity is closer to a human observer. It would be interesting to know which current cameras best fulfil this ideal :-) Such a camera would capture "what you see" with much less need for manual post capture tweaking. Graeme Gill.
Hi to all. There is a fundamental difference between colorimetry and appearance. There is also a common problem with colorimetry itself. The standard observer functions are based on averages, there are 10 degree and 2 degree models that are different. Naturally there are different illumination/collection geometries. We often see displays with different backlight technologies measure exactly the same, but when viewed by a single observer they appear quite different. We don't have strong science in the area of adaptation in emissive/reflective environments. Colorimetry and Spectrophotometry are very good for measuring color, but that doesn't mean that the result of the measurement tracks the appearance of the objects. The argument of colorimetric accuracy in the field has very little to do with the quality of the output with respect to appearance. On the other hand, you do need some notion of the colorimetry to apply appearance principles, but those principles modify the "colorimetry" to the point that the notion of "accuracy" is really not much of a factor. At the ICC, we are looking at some of these issues to improve the total color appearance transfer of data from input to output. http://www.color.org/icclabs.xalter This effort is aimed at multispectral imaging chains and advanced imaging work chains that carry multidimensional reflectance models. This work is going to be an open source effort and we will be building a reference implementation as well. Regards, Tom Lianza Co Chair ICC . On 6/7/13 10:06 AM, "Henry Davis" <davishr@bellsouth.net> wrote:
Thank you, that's interesting and prompts me to ask: do colorimeters/ spectrophotometers attempt to mimic the human observer?
I suppose there can only be a claim for mimicking - for how could it ever be verified? But we use them for measuring and determining a colorimetric match so we have to rely on the claim because besides that there is only subjective observers. Or we rely on both.
My question about a measured match between two samples that don't match visually is still important to me and I think the answer puts a lot of stress on how colorimetric accuracy is defined.
Henry
Graeme wrote:
Henry Davis wrote:
Nobody is saying that the illuminant isn't a hugely important factor.
The main reason that the illuminant is a factor is that many/most cameras aren't colorimetric. By that I mean that their spectral sensitivities don't match a human observer. If they did, and their (light level) linearity was known, then (by definition) they'd be a high spatial resolution colorimeter.
The practical situation is analogous to the one with cheap display colorimeters: They can be very accurate with a correction matrix if the spectral nature of the samples they are reading is known. With a display device that has colors that are (largely) a liner combination of the three primary spectra, a matrix does nicely. For a camera, it can't be that easy, because real world object colors are composed of many metamers, and the spectrum hitting the camera sensor is that times the illuminant spectrum. So a "calibration" matrix can fudge it for a range of expected object spectra x expected illuminant, but ultimately it's going to work the best over the widest range of conditions if the camera spectral sensitivity is closer to a human observer.
It would be interesting to know which current cameras best fulfil this ideal :-)
Such a camera would capture "what you see" with much less need for manual post capture tweaking.
Graeme Gill. _______________________________________________ Do not post admin requests to the list. They will be ignored. Colorsync-users mailing list (Colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/colorsync-users/tlianza%40xrite.co m
This email sent to tlianza@xrite.com
Tom, You think that some day soon (a few years?), i1profiler will incorporate a spectral imaging model? New, yet to be drafted ICC specs will document the required color transforms? I suppose X-Rite like Kodak and a few other manufacturers will boast their "proprietary" spectral imaging implementations. 2016? Best / Roger -----Original Message----- From: colorsync-users-bounces+graxx=videotron.ca@lists.apple.com [mailto:colorsync-users-bounces+graxx=videotron.ca@lists.apple.com] On Behalf Of Thomas Lianza Sent: Friday, June 7, 2013 11:14 AM To: Henry Davis; colorsync-users@lists.apple.com Subject: Re: Colorimetric Accuracy in the Field Hi to all. There is a fundamental difference between colorimetry and appearance. There is also a common problem with colorimetry itself. The standard observer functions are based on averages, there are 10 degree and 2 degree models that are different. Naturally there are different illumination/collection geometries. We often see displays with different backlight technologies measure exactly the same, but when viewed by a single observer they appear quite different. We don't have strong science in the area of adaptation in emissive/reflective environments. Colorimetry and Spectrophotometry are very good for measuring color, but that doesn't mean that the result of the measurement tracks the appearance of the objects. The argument of colorimetric accuracy in the field has very little to do with the quality of the output with respect to appearance. On the other hand, you do need some notion of the colorimetry to apply appearance principles, but those principles modify the "colorimetry" to the point that the notion of "accuracy" is really not much of a factor. At the ICC, we are looking at some of these issues to improve the total color appearance transfer of data from input to output. http://www.color.org/icclabs.xalter This effort is aimed at multispectral imaging chains and advanced imaging work chains that carry multidimensional reflectance models. This work is going to be an open source effort and we will be building a reference implementation as well. Regards, Tom Lianza Co Chair ICC .
I am desperately trying to keep this work Open source with a reference implementation for developers to use. There will be a presentation next week by Max D. who is doing most of the spectral work. One of our X-rite team is working on the appearance aspects.I think we will open the project up before the end of the year. I really am pushing to keep the "proprietary" stuf to a minimum. This is so complex that it needs real working examples. The documentation will probably take a year, but it has already started. I will come back with a report after the ICC meeting next week. On 6/8/13 2:49 PM, "Roger Breton" <graxx@videotron.ca> wrote:
Tom,
You think that some day soon (a few years?), i1profiler will incorporate a spectral imaging model? New, yet to be drafted ICC specs will document the required color transforms? I suppose X-Rite like Kodak and a few other manufacturers will boast their "proprietary" spectral imaging implementations.
2016?
Best / Roger
-----Original Message----- From: colorsync-users-bounces+graxx=videotron.ca@lists.apple.com [mailto:colorsync-users-bounces+graxx=videotron.ca@lists.apple.com] On Behalf Of Thomas Lianza Sent: Friday, June 7, 2013 11:14 AM To: Henry Davis; colorsync-users@lists.apple.com Subject: Re: Colorimetric Accuracy in the Field
Hi to all.
There is a fundamental difference between colorimetry and appearance. There is also a common problem with colorimetry itself. The standard observer functions are based on averages, there are 10 degree and 2 degree models that are different. Naturally there are different illumination/collection geometries. We often see displays with different backlight technologies measure exactly the same, but when viewed by a single observer they appear quite different. We don't have strong science in the area of adaptation in emissive/reflective environments. Colorimetry and Spectrophotometry are very good for measuring color, but that doesn't mean that the result of the measurement tracks the appearance of the objects.
The argument of colorimetric accuracy in the field has very little to do with the quality of the output with respect to appearance. On the other hand, you do need some notion of the colorimetry to apply appearance principles, but those principles modify the "colorimetry" to the point that the notion of "accuracy" is really not much of a factor.
At the ICC, we are looking at some of these issues to improve the total color appearance transfer of data from input to output. http://www.color.org/icclabs.xalter This effort is aimed at multispectral imaging chains and advanced imaging work chains that carry multidimensional reflectance models. This work is going to be an open source effort and we will be building a reference implementation as well.
Regards, Tom Lianza Co Chair ICC .
Henry Davis wrote:
Thank you, that's interesting and prompts me to ask: do colorimeters/spectrophotometers attempt to mimic the human observer?
By definition that's what a colorimeter sets out to do - to match the standardised human color matching curves (CIE). A spectrometer can do whatever you want - you can mimic all sorts of different human and non human observers who sense light over the visible range of wavelengths. Graeme Gill.
participants (4)
-
Graeme Gill
-
Henry Davis
-
Roger Breton
-
Thomas Lianza