Questions of color in digital photography. Do all matter in a magical slider WB?


16.10.2008 0:03:00
Due to the increased interest in the issue of the offer is not quite appropriate for this discussion threads about Nikon D700 move here. Well, at the same time all the issues tied to the difference SSD KMOS can be discussed here.

Start the discussion can be read here, rewind slowly five pages ago.
Nikon D700 Volume II. # 158




Mini guide on



Useful links within topics:
  • «Crossing Europe" on registration of the visible spectrum and the criterion Luther-Ives. In a very (very!) Compressed form explains why the results of shooting the same scene by different cameras often can not be reduced to one color
  • concept of "visual stimulus" and "metameric»
  • about the "locus". The destruction of the myth of Ultra Wide coverage WideGamut RGB and other color spaces with the Cardinals is the locus
  • link on the spectral characteristics of some digital cameras Use direct link
  • brief report on the study of the spectral characteristic of a 9 digital cameras. Link to ifolder. ru have probably outdated. New link - http: // files. unifiedcolor. com / forWeb / Camera sensitivities. zip
  • denial of the possibility of WB (White Balance) in the general case
  • confirmation of the possibility of creating chambers (almost) arbitrary spectral characteristics of the sensor
  • again the impossibility of chromatic adaptation in general
  • refutation of the legend of the "red peak in the blue range." This peak is not!
  • concept of "sensitive description of the stimulus" and "constructive description of the stimulus." The theorem of classification descriptions. Additional explanations - and trihromatizm Grassmann. Transformation formulas between sensitive and constructive description
  • Link to very useful site. Distribute tabulated functions, one way or another related to vision, image science, and so on. D.
  • of constancy of the spectral characteristics of the retina
  • very useful about the color gamut, sRGB, Adobe RGB, WideGamut RGB, and so on. D.
  • link to a useful discussion of "Sight - a comparison of how the principle of the device. " Mechanisms of primary processing of visual apparatus trichromats values ​​
  • white balance
  • of populations of retinal cones
  • depending on JND (just noticable difference) on various factors on the impact
  • filters on metamerism (color discrimination ability) view (with the calculations, numbers and pictures )
  • of the agreement on the use of the terms "white balance" and "chromatic adaptation»
 

16.10.2008 1:46:00
http: // vgrin-photo. front. ru / cmos_ccd / view. jpg picture piqued my interest in the subject
http: // vgrin-photo. front. ru / cmos_ccd / my vkladd in iisledovanie question

Two CCD: Nikon D70 and Fuji S5Pro, the lens is the same, the time gap between shots ~ 10 seconds.
skintone more or less the same, but the greens are very different http: // vgrin-photo. front. ru / cmos_ccd / s5pro_d70 / s5pro_d70. jpg
And it's not in the profiles, the picture http: // vgrin-photo. front. ru / cmos_ccd / s5pro_d70 / s5pro_d70_AdobeRGB. jpg applied to both images AdobeRGB, as a camera profile.

This image collected from the color channels RAW, that is, without any interpolation konverorah, so to say "fair colors." To the red and blue channels applied level, only the right engine, ie multiplication coefficients for some, to obtain the desired BB. And to all isobrazheniyu applied gamma 2. 2. Appointed profiles AdobeRGB and all converted into sRGB. In pictures http: // vgrin-photo. front. ru / cmos_ccd / s5pro_d70 / s5pro ... _AdobeRGB_raw. jpg BB made by the skin tones in http: // vgrin-photo. front. ru / cmos_ccd / s5pro_d70 / s5pro ... eRGB_raw_grey. jpg on the wall under the right ear - she sdelalna "gray". In both cases, the greens are very different

State of CCD / CMOS for 2003 (in English) Charge coupled CMOS and hybrid detector arrays

16.10.2008 2:52:00



Indeed, and what have the BB?

IMHO, has been repeatedly discussed.
Bayerova matrix consists of three important components:
1. AA filter - to avoid moire related to the structure of the matrix.
2. filters to select the right color - usually RGGB.
3. recording element - CMOS TSTSD.

1 - is appartany interpoyalyator as if affects the sharpness.
3 - is DD noise. . . sensitivity. . . little effect on the color, ie. a. a different spectrum can be recorded in different ways.
2 - mainly responsible for the color reproduction.

BB here as a little bit to do with it, that is. To. He Responsibility for what is considered gray. . . but not for the flowers themselves - RSL and their preparation of the matrix.

By tsvetofilram. . . I do not really understand why there are from calculations, but it is believed:
1). If they make a very narrow, it is not good. . . t. To. They should at least cover the entire visible spectrum - this time (ie. some color to an otherwise principle can not be registered). Two - is that the sum would be better IMHO to have bandwidth = 1,. E. That there is no loss of energy for heating the filter.
2). Principle, if everything was perfectly adjusted would be sufficient to have only two hours used display. 1st - a monotonic increase in capacity with respect to wavelength. 2nd vice versa. Then the signal with respect to the matrix of the adjacent cells can be found wavelength. Well, her energy.
You can simplify the system by placing two filters before the matrix, which would cut off a man invisible spectrum.


In ordinary matrices used the three primary colors, t. To. It is associated with human vision and perception. With this achieved good accuracy in determining osnovnh colors - t. E. Particularly blue, green, red. . . . but the accuracy of the determination of intermediate colors - suffers.

Well. . and then every vendor offers its own implementation , but again IMHO theory seems to be enough just two filters, if everything was perfect.

Maybe nonsense written. . . do not hurt much.


update:
understood that nonsense written) crossed out. IMHO, it would be better to curves filters as close as possible to the curves approaching the perception of eye different wavelengths, or shoals will be inevitable.

16.10.2008 3:36:00
Hmm. . . all the specifics of color should be considered in conjunction with the specific color converters and profiles cont separate chambers.

16.10.2008 10:54:00

16.10.2008 10:59:00
quote:
:
Here is my version of the treatment, with reduction to one color:
http: // s54. radikal. ru / i144 / 0810 / 9a / 17223090cee0. jpg (996x750, 216, 9Kb)
http: // s54. radikal. ru / i144 / 0810 / 9a / 17223090cee0. jpg

color quite different. To the left is much better.

16.10.2008 11:08:00
quote:
:
color quite different. To the left is much better.

As for "much" is what you have got excited. Besides, I did not aim to do better or worse, but simply brought the two halves of the image to approximately the same, and is suitable for printing color.
Waiting for your treatment options.

16.10.2008 11:11:00
quote:
: What about "more" is what you have got excited. Besides, I did not aim to do better or worse, but simply brought the two halves of the image to approximately the same, and is suitable for printing color.

What can I say? You did not happen. Absolutely. The colors are different, the reception left much estestvennno.

my treatment option there is no need to wait, because I'm of the opinion that the colors are CMOS / CCD and color cameras from different manufacturers is different, and it happens that much. And lead to the same color images will not work.

16.10.2008 11:18:00

As for "much" is what you have got excited. Besides, I did not aim to do better or worse, but simply brought the two halves of the image to approximately the same, and is suitable for printing color.
Yes, you are right that's no good, gray wall has become pink, and the skin was like a pig.

16.10.2008 11:18:00


as I am of the opinion that the colors are CMOS / CCD and color cameras from different manufacturers differ,
I would like to hear your opinion about the differences in color reproduction Canon / Nikon. On CMOS.



Two - this is a total IMHO it would be better to have the bandwidth = 1
= 100% efficiency = a perpetual motion machine.

16.10.2008 11:52:00
Skintone with the D700, after processing
left to the original, to the right treatment

http: // s45. radikal. ru / i107 / 0810 / c9 / d488b2e19c55. jpg

Removed pinkish otttenok, negritoska became Chornaja. Dress too.


16.10.2008 12:05:00
quote:
:

Yes, you are right that's no good, gray wall has become pink, and the skin was like a pig. But

hair black steel, not reddish.

And boy do any color can be made, though green:

get 10otlichy:
http: // i069. radikal. ru / 0810/05 / 7d4d6aedfb79. jpg

Takchto, all garbage, FSH taxis, and if you think your camera unutre no photoshop. . . . and RAW files you get rough, especially TSMOS where all processing is on the chip. . . . it's time to dispel your naïve delusion!


16.10.2008 12:18:00

= 100% efficiency = a perpetual motion machine.
I meant that if f x (l) - a graph passing filter X depending on the wavelength l, then for all wavelengths in the visible spectrum f r (l) + f g (l) + f b (l) = 1

16.10.2008 12:23:00

In dark-skinned people in the yellow color of the skin may be a third more than the purple, and the share of blue may be one-third of the share of purple. It's all roughly, of course. If you can work in CMYK, of course.

16.10.2008 12:30:00
quote:
:

get 10otlichy:
http: // i069. radikal. ru / 0810/05 / 7d4d6aedfb79. jpg

I would put it differently: get 10 matches. The colors are different. I wish I could be like you and do not notice it.

16.10.2008 12:39:00
quote:
:
in dark-skinned people in the yellow color of the skin may be a third more than the purple, and the share of blue may be one-third of the share of purple. It's all roughly, of course. If you can work in CMYK, of course.


By the way, does anyone met programmu- "dropper", which would indicate "fotoshopovskie" of CMYK? . The fact that available to me the program - "pipette" show an abstract CMYK, such as web colour picker shows the color
0C 45M 50Y 0K and photoshop 5C 50M 60Y 0K. Which of these cmyk "more correct"? And whether the pipette, the figures show a la Photoshop?

16.10.2008 12:40:00
Not
not garbage. IMHO the case still in the filters of the RSL in the matrix. If they are bad, the color is not vytsepit.

16.10.2008 12:41:00

get 10otlichy:
http: // i069. radikal. ru / 0810/05 / 7d4d6aedfb79. jpg

Excuse me, but how right he was pink, and has remained, but became a little less. One of my sign tsvetokorretor a complexion calls pig and sends me here such Smal:

But he owns Photoshop well, so it kmos / CCD little concern. I have the same armless blue, green, pink skin tone strain.


These photos to one color only adjust the mask, IMHO, can cause. For initial skin tone do the same or similar, and the greens are very different.


16.10.2008 12:49:00
quote:
:

not no garbage. IMHO the case still in the filters of the RSL in the matrix. If they are bad, the color is not vytsepit.

"bad" filter does not exist! Read about Foveon, that's really where complete w. . . filters - they are not at all, in fact, tsvetopogloschenie matrix layers overlap, and anything with the help of simple linear algebra restore the original color, so much so that images with foveona - just a feast for the eyes, the Bayer color in comparison looks flat and dull.

http: // www. ixbt. com / digimage / foveon2. shtml

16.10.2008 12:52:00

Takchto, all garbage, FSH taxis,
Judging from the above examples - can and taxis, but not in your hands for sure. Skintone to the right on the left below the average, and of the differences are obvious. . .

16.10.2008 12:55:00


good stuff for. . . April 16, 2002 Mr.

16.10.2008 12:56:00
quote:
:
These photos to one color only adjust the mask, IMHO, can cause. For initial skin tone do the same or similar, and the greens are very different.
Tov. Margulis teaches us that it is wrong - the boy fall out from the background.

quote:
:

Takchto, all garbage, FSH taxis,
Judging from the above examples - can and taxis, but not in your hands for sure. Skintone to the right on the left below the average, and of the differences are obvious. . .

"It's not Photoshop........." 'Said Wolf and blushed. . . . .

16.10.2008 13:00:00

Tov. Margulis teaches us that it is wrong - the boy fall out from the background.
A friend Margulis taught to do from malnika pig?

16.10.2008 13:02:00
quote:
:
A friend Margulis taught to do from malnika pig?

And you print my photos, and donate the boy - he will grunt of happiness!

16.10.2008 13:03:00

after sorting out all the profiles the harmony!

16.10.2008 13:10:00

I read the article, but some things do not really understand. In any case, there is a lot like a little glossed over. Well, anyway, I do not understand why the color definition in faveon better than Bayer Leverkusen.
 

16.10.2008 13:13:00

"It's not Photoshop........." 'Said Wolf and blushed. . . . .

"It's not skintone boy on the lips face........" - Little Red Riding Hood said, and blushed even thicker. . . . . . . . .

16.10.2008 13:15:00
quote:
:
get 10otlichy:
http: // i069. radikal. ru / 0810/05 / 7d4d6aedfb79. jpg

There they were not 10. There they were much more. Completely different in color images.

quote:
:
Tov. Margulis teaches us that it is wrong - the boy fall out from the background.

I realized that you are color blind, which he tried to teach the right color!

16.10.2008 13:18:00
quote:
:
. . . . it's time to dispel your naïve delusion!
color is different and strong. then, if I have 6-8 gig with different cameras, how to be, you send?

16.10.2008 13:22:00
quote:
:

Maybe nonsense written. . . do not hurt much.

update:
understood that nonsense written) crossed out. IMHO, it would be better to curves filters as close as possible to the curves approaching the perception of eye different wavelengths, or shoals will be inevitable.

I'm certainly not a pro in color reproduction, but I think all your arguments about the color is incorrect.

I understand that in order to maximize coverage of narrow-band filters are needed, but it is contrary to the requirements of low noise.
How technology affects color reproduction photodetector I do not really understand. Probably, CCD and CMOS have different spectral sensitivity characteristics, and this is taken into account in the design of filters.

At present the problem of the color reproduction of different technologies I refer rather to build the correct profile that describes the device registration.
And certainly worsen the situation with coverage extension band filters may be in pursuit of a low noise, but it seems to me that proper to the CCD / CMOS is not relevant.

16.10.2008 13:56:00

here as though there are two approaches.

first - is to get the maximum color gamut. Then yes. . . it was necessary to filter as much as possible and already, but it is necessary that any wavelength of the visible color of its recorded at least one filter. Otherwise, get that for the wavelength 500 nm it will not register and no filter. . . as if the image will be. So you can not make a little filters and that they are narrow.

you is probably confused with the radiation of the visible spectrum - there yes. It is necessary that the radiators were as narrow as possible range, regardless of their number - at least there will be two. . . 22. Better least to emit only one wavelength. But then again. . . This is true especially for human perception, and not for an abstract of the emitted spectrum.

the other hand. We have three filters with the characteristics of the cones in the eye - t. E. Their curves coincide. We have three emitter with a spectrum and its intensity = filter characteristics and cones in the eye.

using filters and matrix register range. There is a loss of information. . . some of the colors we do not take my distinguished - consider wavelengths. Continue to generate the spectrum using izluchaeteley according we registered. The spectrum will be different from the original. . . but the man did not notice. . . t. To. gamut of our receiver and transmitter is color gamut eyes. For instruments will it be a different spectrum. . . for flies can be too. . . . or bees. . . someone there to reach more?


Probably, CCD and CMOS have different spectral sensitivity characteristics, and this is taken into account in the design of filters. Yes
not have. . . TSTSD and CMOS - a different system of organization of cells. The registration process there seems to be one. . . and depends on the type of silicon.

16.10.2008 14:00:00

I understand that in order to maximize coverage of narrow-band filters are needed, but it is contrary to the requirements of low noise.

Actually, to achieve high fidelity color filters in front of the Bayer matrix not need to be narrowband. If you delve into the basics of color, then arrive at the opinion that the RGB color space is ambiguous when registering continuous spectrum of electromagnetic radiation, and even the amount of monochromatic radiation. The only reason why RGB working - is the fact that RGB - simulates system tsvetoregistratsii human eye. Accordingly, the only way to obtain a high color fidelity - is to create one RGB filter characteristics as close as possible to the characteristics of the sensitivity of the human eye retsetorov. Variations on a theme of narrowing or expanding bandwidth RGB filters inapplicable lead to distortion of color reproduction.

16.10.2008 14:02:00

Here is my version of the treatment, with the reduction to one color:
turned green skin, hair and so on.


and lead to the same color images will not work.
+1. same story when you try to bring the figure to the film, it seems sometimes, and sometimes does well http: // vgrin2. front. ru / Leica /


And boy do any color can be made, though green:
do not hurt my baby

0C 45M 50Y 0K and photoshop 5C 50M 60Y 0K. Which of these cmyk "more correct"? And whether the pipette, the figures show a la Photoshop?
it depends on the CMYK profile, because it can be configured as you like, the "right" CMYK does not exist. All recommendations Margulis tied to his proposed book early setting of default CMYK

And you print my photos, and donate the boy - he will grunt of happiness!
unlikely he had seen normal


I understand that in order to maximize coverage of narrow-band filters need
not. if so do the blue will fall between blue and green and turn black. Bands overlap and each color must match the combination of the three (or more depending on the implementation) signals. Many interesting things can be found here, if you have the strength. The main criteria for accurate color reproduction systems

p. s. added in the second post a link to an interesting article

16.10.2008 14:03:00
Yes
not have. . . TSTSD and CMOS - a different system of organization of cells. The registration process there seems to be one. . . and depends on the type of silicon.

Actually even transistors to create one by the same technology (CMOS or CCD) may have different spectral sensitivity Much depends on the details. . . .

16.10.2008 14:09:00
Yes
not have. . . TSTSD and CMOS - a different system of organization of cells. The registration process there seems to be one. . . and depends on the type of silicon.
there are nuances http: // vgrin-photo. front. ru / cmos_ccd / Janesick_paper_2003. pdf

16.10.2008 14:17:00

color formed in the brain of the user example of this color illusion:

http: // flogiston. ru / blog / color_illusion
http: // fun. nashcat. ru / illusions / contrast. php? page = 18
is significant about the "right" colors and skintones:
http: // fun. nashcat. ru / illusions / contrast. php? page = 17

16.10.2008 14:26:00

something I probably did not understand. There are silicon - it registers photons. . . napruga on it appears. There is a capacitor or something like that, that razyazhaetsya. Then the charge was withdrawn. . . and considered. Where he is removed as removed as passed. . . it all depends on the CMOS TSTSD. But the color. . . where does the color. Noise - yes, razyne.


Just will not do. and short?


There is an opinion that the noise can be different - color. . . BW (luminance)?
There may be a difference between CMOS TSTSD? I now think that may be. Luminance noise - a noise that happens when simultaneously strengthen all three colors in the sensor - well, aiming Suppose something. A color (or color) - is an individual channel noise. . . t. e. a single sensor. That Is Why. . . mei bi and depends on the organization - Kaka organization which produces noise.

But then again. . . D80 gives a zhpege luminance noise and A100 - color. IMHO the processor is also possible to correct one another.

16.10.2008 14:31:00

Just will not do. and short?
most interesting for us
1. CCD AND CMOS IMAGER DEFICIENCIES

16.10.2008 14:38:00

Well, that's understandable. . . More interesting brown color in nature.


There are several levels of perception. . . on one of the damage occurs automatically adjust BB.

16.10.2008 15:08:00

color formed in the brain of the user example of this color illusion:
MDA, funny stavnivat malotsvetnye illusion photo.

16.10.2008 15:16:00

ran fluently. . . I understand that it is different structure of sensors still generates slightly different methods of registration?

16.10.2008 15:17:00

something I probably did not understand. There are silicon - it registers photons. . . napruga on it appears. There is a capacitor or something like that, that razyazhaetsya. Then the charge was withdrawn. . . and considered. Where he is removed as removed as passed. . . it all depends on the CMOS TSTSD. But the color. . . where does the color.

Ha, ha, ha. . the most basic example for you - Foveon. There are light-sensitive elements have a relatively low sensitivity range - t. E. Roughly speaking for different wavelengths (different photon energies) the efficiency of translation in "naprugu" can vary significantly.

Accordingly, if the spectral sensitivity of CCD and CMOS are different, the characteristics of the filter for accurate color reproduction on a CMOS and CCD are different.

16.10.2008 15:42:00

Accordingly, if the spectral sensitivity of CCD and CMOS are different, the characteristics of the filter for accurate color reproduction on a CMOS and CCD are different.
I do not understand. I thought that no one and nothing prevents to TSTSD and put the same CMOS silicon. In an article that led I realized that just says that in the way of registration requirements and change it. that can be put in TSTSD was impossible to put in CMOS.

t. E. The same is it that different silicon CMOS and TSTSD?

But then again. . . scanned. . . Many forms and more. No mood cut one.

16.10.2008 22:42:00

I do not understand. I thought that nobody and nothing prevents to TSTSD and put the same CMOS silicon .

Of course you did not understand me, t. To. Apparently have a very poor knowledge in the field of semiconductors. I'm telling you that even within the same technology can be the difference in the spectral sensitivity. Any semiconductor device - it kompromisny set properties. What would you have understood: for example, you can create one car is small and nimble or more but clumsy. Do the same and more nimble and nepoluchaetsja. Although all put the same steel. Now think about the different technologies - such as elektrovoz and locomotive - no one bothers to put them Odinokova steel , but the dynamic characteristics they are very different.

So, when we start to move from one technology to another, the optimum properties of the displaced. So we get that quite sure there is a difference in spectral sensitivity between CCD and CMOS, and the Problem not in silicon, but the fact that constructively make CCD and CMOS sensors with the same parameters is difficult / expensive or impossible. Although silicon is everywhere in fact one and the same.

16.10.2008 23:14:00

Pasib. grasped.

17.10.2008 0:39:00
Thank you raised an interesting topic!

Quick naive presentation on the theoretical minimum.

1) Camera - measuring instrument that registers some parameters EMR focused (incident on) the light-sensitive sensor (matrix).

2) EMI (electro-magnetic radiation) is characterized by a function of the spectral distribution of the photons (or energy, depending on the application, in this case it is convenient photons). Further use for the spectral distribution of photons F (l), G (l) and r. D.

3) The matrix is ​​composed of a plurality of photodetectors, each of which detects (measures) a single aspect (a single number) of the physical phenomenon of "incident on the sensor light. "

4) Each photodetector is characterized by the spectral sensitivity function p (l) - the probability of detecting a photon with a wavelength l. The integral of the product of p (l) * F (l) is proportional to the measured detector per unit time value when focusing on it (the detector) EMR with spectral photon distribution function F (l). Exposure time integral of the integral of the product of p (l) * F (l) are measured detector value. Notation: r (l), g (l), b (l) - the spectral sensitivity of the detectors; R, G, B - the measured values ​​of the detectors. Next, we consider only "tricolor" matrix. The generalization to an arbitrary n-dimensional case is constructed similarly.

5) For simplicity we consider the case that the spectral distribution of photons F (l) is constant in time and across the entire field of view optical system. Then, obviously, the matrix is ​​a linear operator on the island of plural spectral distributions of photons in the first (three nonnegative) quadrant of the three-dimensional govt R 3 .

6) Two AMY F 1 (l) and F 2 (l) are indistinguishable to the system of three photodetectors r (l), g (l), b (l), if R 1 = R 2 , G 1 = G 2 and B 1 = B 2 . Obviously, for any EMR there are infinitely many indistinguishable him this system detectors other EMR.

7) There is a special system of three sensors - a human. Obviously, two EMP distinguishable person is not always distinguishable trёhsensornoy other system and vice versa.

8) If (omitting a lot of text) consider linear embedding R 3 , in a sense opposite to the linear operator defined in (5) (here omitted really many unobvious text), the composition of the operator (5), and this attachments are the orthogonal projection on a three-dimensional subspace of the linear (infinite-dimensional!), made in all EMR. Obviously, the three-dimensional subspaces beskonechenomernogo space infinite. Consequently, there may be any number of different three-component systems of registration EMR.

9) It is obvious that to calculate the components R, G, B in the same three-component system based on measurements other three-component system of registration of EMR can be if and only if there are attachments R 3 in space vsёh EMR that the composition defined in ( 8) are identical. In other words - when both systems describe the same three-dimensional subspace of all EMR.

10) Adoption (9) known in colorimetry as a criterion for Luther-Ives (Luther-Ives, 1914). The wording (modernized) - to be able to colorimetrically accurate color registration ternary system photodetectors is necessary and sufficient that its spectral sensitivity photodetectors r (l), g (l) and b (l) are linear combinations of the spectral sensitivity of cones standard observer.

11) required (and sufficient) condition for registration of colors soon known as 100 years. Why is the issue of compliance of the spectral characteristics of modern photodetectors criterion Luther-Ives neglected? I do not know. Why so much "walk" the spectral characteristics of matrices? I do not know.

12) destruction myth №1 . "For accurate color need to build a profile for a large set of samples with known colors." This is not true. Failure to comply with the criterion of Luther-Ives accurate color possible. At all. No computer model (or linear or polynomial or CLUT, no other) can not fix a simple fact - if out of criteria Luther-Ives there are infinitely many EMR distinct person and not distinguishable system photodetectors, and vice versa - is not distinguishable and recognizable human system photodetectors. Once again - if your camera does not meet the criteria Luther-Ives (and it is almost certain - almost, because we have not explored all ever issued by the camera - so) accurate color possible. In view of the fundamental physical reasons. Law of nature.

13) destruction myth №2 . "The color gamut of the camera (scanner)." The concept of the gamut of photodetectors pops up in discussions frequently. And this despite the fact that, obviously, in the registration of any EMR any system (three) photodetectors regularly gives (three) measured values. Once again - any EMI recorded by any system of photodetectors. Thus, any camera detects any radiation. If the camera does not distinguish between two distinct human EMR, then we are not talking about the impossibility of registration of one of them, but only an abuse criterion Luther-Ives and therefore of the illegality of the applicability of the word "color" to the camera at all.

Links:
* Special topic - Basic Criteria accurate color reproduction systems (although the theme of "plump" ferret out there useful information quite difficult)
* presentation format PPS (PowerPoint Show) - http: // www. kweii. com / site / color_theory / 2007_SJ / Quali ... lityCriterion. pps
* outline format HTML - http: // kweii. com / site / color_theory / cri_ru / cri_introduction_ru. html

upd1
Notes to two points (and, at the same time, a simplified exposition of them):

8 ") This is a three-dimensional subspace Gunter Wyszecki (G. Wyszecki) proposed to call" the fundamental color space "in the case of the" human "of photodetectors. In view of (6), obviously, each of the photodetectors, there are infinitely many fundamental color spaces. The problem of choosing one of them (the "most correct") put in his doctoral dissertation. The first approximation to a possible solution, useful in practical problems, described in monograph Joseph Cohen (Jozef B. Cohen, Visual Color And Color Mixture, University of Illinois Press, 2001).

9 ") This statement can be formulated easier, namely to calculate the components R, G, B in the same three-component system according to Measurement of other ternary system of registration of EMR can be if and only if (any) fundamental color space of a system of detectors is a fundamental color space of the other.

Attention! Control warning. I recall that the color - it is the attribute of human visual perception. In connection with this use the adjective "color" in the preceding two paragraphs, it is inappropriate and only allowed only for reducing the volume of text.

upd2
Closer to the topic . Summary of the "theoretical minimum".

Results playing the same scene shot with different cameras from different manufacturers vary in color for the simple reason that different cameras recorded different physical characteristics EMR. Usually (if not always) have not to say that a lot in common with helluva proper color.

quote:
:
the other hand. We have three filters with the characteristics of the cones in the eye - t. E. Their curves coincide. We have three emitter with a spectrum and its intensity = filter characteristics and cones in the eye.

using filters and matrix register range. There is a loss of information. . . some of the colors we do not take my distinguished - consider wavelengths. Continue to generate the spectrum using izluchaeteley according we registered. The spectrum will be different from the original. . . but the man did not notice. . .
Wrong. A man will almost certainly notice a change of tone.

Make this easy. Enough, for example, to calculate the response of the human eye with the spectral distribution of the radiation energy is numerically (very important word! Note) equal sensitivity of one of cones. I did this for all three cones, using data Stockman-Sharpe (2000 standard 2 degree observer CIE) - http: // ifolder. ru / 8608595

As you can see, EMI with a spectral distribution, abundance (very important word! Note) sensitivity equal to any of the cones stimulates response all three cones.

If we denote by L, M, S cone responses for some EMR and creating a system of radiators that you have described, sgenenriruem radiation with a spectral distribution numerically (for the third time to draw your attention - it is numerically very important word!) equal to L * L (λ) + M * M (λ) + S * S (λ) (where L (λ), M (λ) and S (λ) - spectral sensitivity cones) , then the response of the human eye, this radiation will obviously be proportional to (2 59 * L + 2. 08 * M + 0. 13 * S; 2. 08 * L + 2. 00 * M + 0. 21 * S; 0. 13 * L + 0. 21 * M + 1. 25 * S). This vector is, obviously, not necessarily collinear with the initial vector (L, M, S).

as required.

Important note about the number .
Above I made note of three of the importance of the word "numerical" in the expression "numerically equal." In fact, it should be understood that for the numbers used for the purposes of measurement, storage and playback of certain aspects of the physical phenomena are very specific physical quantities. Actually the number of physical problems, as a rule, not the number, but dimensional quantities . Numbers can always compare. Physical quantities - only in the case of coincidence of the physical dimensions and units. For example, the number of 9. 81 may indicate the approximate value of the acceleration of gravity 9. 81m / s 2 , but can - a lot of camp layout in the "budget" a two-week trek 9. 81kg. Of course, 9. 81 = 9. 81 - right. 9. However, 81m / s 2 = 9. 81kg - not very (or hate) good score in physics.

draw your attention to the fact that the spectral sensitivity and spectral energy distribution (or photons) - various physical quantities. Hence, equating (and even direct comparison!) Spectral characteristics of emitters and sensors - rough physical error. So to do.

17.10.2008 14:21:00
:

It seems to me, is what I waited a long time in discussions on different color reproduction devices, let me make a squeeze:

quote:

9) It is obvious that to calculate the components R, G, B in a three-component system based on measurements other three-component system of registration of EMR can be if and only if there are attachments R 3 in space vsёh EMR that the composition defined in (8) are identical. In other words - when both systems describe the same three-dimensional subspace of all EMR.

10) Adoption (9) known in colorimetry as a criterion for Luther-Ives (Luther-Ives, 1914). The wording (modernized) - to be able to colorimetrically accurate color registration ternary system photodetectors is necessary and sufficient that its spectral sensitivity photodetectors r (l), g (l) and b (l) are linear combinations of the spectral sensitivity of cones standard observer.

11) required (and sufficient) condition for registration of colors soon known as 100 years. Why is the issue of compliance of the spectral characteristics of modern photodetectors criterion Luther-Ives neglected? I do not know. Why so much "walk" the spectral characteristics of matrices? I do not know.

12) destruction myth №1 . "For accurate color need to build a profile for a large set of samples with known colors." This is not true. Failure to comply with the criterion of Luther-Ives accurate color possible. At all. No computer model (or linear or polynomial or CLUT, no other) can not fix a simple fact - if out of criteria Luther-Ives there are infinitely many EMR distinct person and not distinguishable system photodetectors, and vice versa - is not distinguishable and recognizable human system photodetectors. Once again - if your camera does not meet the criteria Luther-Ives (and it is almost certain - almost, because we have not explored all ever issued by the camera - so) accurate color possible. In view of the fundamental physical reasons. Law of nature.

13) destruction myth №2 . "The color gamut of the camera (scanner)." The concept of the gamut of photodetectors pops up in discussions frequently. And this despite the fact that, obviously, in the registration of any EMR any system (three) photodetectors regularly gives (three) measured values. Once again - any EMI recorded by any system of photodetectors. Thus, any camera detects any radiation. If the camera does not distinguish between two distinct human EMR, then we are not talking about the impossibility of registration of one of them, but only an abuse criterion Luther-Ives and therefore of the illegality of the applicability of the word "color" to the camera at all.

Closer to the topic . Summary of the "theoretical minimum".

Results playing the same scene shot with different cameras from different manufacturers vary in color for the simple reason that different cameras recorded different physical characteristics EMR. Usually (if not always) have not to say that a lot in common with helluva proper color.
Apparently, the main reason for the observed variation in the nature of the different color matrices - is ineligible Luther-Ives. And here is the unfortunate family - all unhappy in its own way becomes clear why the construction of the profile of the issue can not be solved - the information is lost at the stage of registration. . .

PS By the way, can be checked in the "home" conditions the degree of mismatch characteristics of this camera to the criterion? Of instrumentation - only EYE-ONE Pro.

17.10.2008 14:52:00

way, and can be checked in the "home" conditions the degree of mismatch characteristics of this camera to the criterion? Of instrumentation - only EYE-ONE Pro.

Hmm, good question, thank you. We need to think.

correctly understand that Eye-One Pro can measure the spectral reflective characteristics of the sample (at least with good 5-10nm, better, of course 1-2nm step)? E., The device can be applied to the sample, press the button, it will shine carefully calibrated at the factory and the lamp will not just three chiselki (XYZ, CIE L * a * b *, or anything else), but that the spectral energy distribution (or photons) reflected light? If so, with a good (calibrated) light can be an experiment.

However, I am afraid, the accuracy is not so hot - will develop errors bulb unit, the measuring unit of the device, the "good" light possible illumination from insufficient premises zashtorivaniya good reflexes from the skin / hairstyles / clothes experimenter / walls, and so on. D.

17.10.2008 15:06:00
:
quote:

correctly understand that Eye-One Pro can measure the spectral reflective characteristics of the sample (at least with good 5-10nm, better, of course 1-2nm step)? E., The device can be applied to the sample, press the button, it will shine carefully calibrated at the factory and the lamp will not just three chiselki (XYZ, CIE L * a * b *, or anything else), but that the spectral energy distribution (or photons) reflected light? If so, with a good (calibrated) light can be an experiment.

However, I am afraid, the accuracy is not so hot - will develop errors bulb unit, the measuring unit of the device, the "good" light possible illumination from insufficient premises zashtorivaniya good reflexes from the skin / hairstyles / clothes experimenter / walls, and so on. D.

"Staff" software Eye-One Pro provides a spectrum with a step of 10 nm, Argyll CMS like can be removed with a range of 3 nm increments. E. It spectrophotometer.
Also, Eye-One Pro is equipped with a calibration stand Gaugeable in vitro model reflecting. This sample has the same number as the device itself, from which I conclude that his reflection spectral characteristics are stored in the device and used to calibrate before measurements reflecting samples, ie. About. compensating error is quite normal incandescent bulbs used to illuminate the surface being measured.
During measurement samples reflecting device is located at a minimum distance from them, and all sorts of extraneous illumination easily deleted.

Also included is diverging nozzle for measuring the spectral characteristics of light.
 

17.10.2008 15:08:00
seems to reproduce colors accurately is necessary not RGB sensors and RGBE well and further, each with a narrower exciting range:

pluses:
increase color gamut and more precise control Composes

cons:
xs as technically

appendage theory " conspiracy "=)

tsmos sensors dragged into the world of digital photography (not the course they just coincidence time) is such a thing as
microlens (all kind of like the original kodak company invented), and now they are gapless, and so on. d.

t. e. on each RGB color filter has a microlens
if she can to make any korretirovki in color? ? ?

same can not be that it is completely optically transparent!

Questions of color in digital photography. Do all matter in a magical slider WB?

info@www.about-digital-photo.com