The middle gray area question, 10-18%

Discussion in 'Mirrorless Digital Cameras' started by randy_smith|10, Jan 25, 2010.

  1. The general understanding that reflective light meters are calibrated to 18% is not quite correct. Canon and Nikon meters are set for 12.5%. Light meter calibration is based on formulas that include a K or C value which is from a large statistical sampling of images shot having better over all exposure from average angles of view of images shot outdoors “under natural disposition of light and shadow” (Ansel Adams, The Negative, Pg 43), a fudge factor if you will. My under standing was that they increase density in the shadow areas, with 12% being about the middle gray calibration center.
    If you shoot an image of a 18% gray card and examine the histogram, sure enough the 12.5% exposure results in a curve to the left of center, about 1/2 stop. One would assume that I would need to add 1/2 stop exposure to reproduce my medium gray 18% back to it’s middle gray value. And the K or C factor was intended to add density, which means more exposure. Seems like our K factor adjustments should result in a middle gray of maybe 22% or so, Higher, not lower than 18%.
    But I did not make any exposure adjustment in my test (again my camera is calibrated for 12.5%), I wanted to see how this exposure of the gray card would look against the middle gray value of Photoshop’s 50% middle gray. Now it seems like the Photoshops 50% middle gray value should be truly middle gray, half way between full white and full black. In fact I shot the Canon 5D mk II, a Gossen Incident light meter, and Minolta spot meter against the same lighting and gray card and compiled the images into a Photoshop image file without exposure adjustment, and compared them all against each other. They were all scary close to each other, less than even a 1/3 stop difference from the 50% middle gray value of Photoshops middle gray value, and all the light meters were slightly darker.
    The only really thing that is clear is, is this really isn’t vary clear. Ansel Adams suggested that without the K factor used the middle gray value would be 18%.
    Is 12 or 13% of one log scale’s middle gray reflectance value able to equal and different scale of printing inks of 18% reflectance, and that maybe we should not be using the word reflectance for one scale or the other? Please mister Wizard! where is middle Gray Really?
    00VbZV-213963584.jpg
     
  2. What Ansel Adams wrote decades ago may or may not 100% accurate for cameras today. Camera design has changed significantly since he died. Additionally not all exposure meters are made the same. Many older cameras out there don't care about color. Some of the newer ones do take color into account. When Ansel Adams started out in photography light meters were not available. Digital sees light differently than film and Ansel Adams never worked with digital. What may be true for Nikon may not be true for Canon or Pentax.
    Additionally what shows up as gray on the camera can look significantly different on the monitor. Before making comparisons between your monitor and printers they should be calibrated so that they all render colors the same.
    For most people "where is middle gray really?" is not as important as using the camera and getting experience with how your particular camera on how it handles various lighting, color, and white ballance situations. Most photographer are more concerned with the final look of the image rather than if the exposure exactly matched the value from a gray card. In fact many people under or overexpose a pictures deliberaely to get the look they want.
     
  3. Actually, the 50% PS gray is not 18% gray in any case--I don't remember the readings-rgb-that are 18% gray, but it is not 127.5. That will be a bit lighter than the 18% gray. And, I would be surprised if the K factor is not still built into meters, as a pro will figure out how to compensate for it and the amateur will want to be sure that Aunt Martha's face is clear, even if backlit! Slightly overexposed images, except with slide film, have always been ok but not so much the other way around.
    The real test, if you are wanting to know how your camera works, would be to shoot a towel, with texture, evenly lit and meter, with the spot mode, and then shoot off frames at the meter reading and then one stop increments in both directions for 4 or 5 stops. Then, put the images in your raw processor and output them with no modification. See where the texture is no longer visible. Then, take the images where the texture disappears and adjust the exposure, in raw, and see again, which one doesn't allow you to recover texture. This will give you your camera's dynamic range at the ISO you shot at. (The digital camera will retain detail longer than it will color, so there is actually a different dynamic range for color than B/W as I have tested it)
    Anyway, once you know that, you start to appreciate what your camera can do or can't.
     
  4. SCL

    SCL

    I guess the question is do you want to know all the technicalities of why one system which really isn't directly compatible with digital works or not under certain circumstances, or do you want your pictures to be predictable and manageable time and time again? I suggest spending more time familiarizing yourself with and mastering the nuances of your particular gear...having an explanation for all the stuff will just drive one nuts.
     
  5. Finally, an interesting question! (Where did they all go?) Your Photoshop 50% gray in RGB is hex #808080, in other words 128 decimal. The 18% gray is 117 decimal [citation, Bart van der Wolf] assuming gamma 2.2 Whereas the Kodak gray card is 160 decimal [same citation, Tony Spadaro] as in the attachment below. This is why many threads say 18% gray is really 12% gray! J.
     
  6. Attachment.
    00VbmX-214110184.jpg
     
  7. Vary true Steve,
    Ansel Adams was vary aware though of the K factor though even in the fifties. Adams also well supported placing exposure were you want for expression, as does any photographer in not too much time in their growth.
    The monitor issue is not relative i think, Photoshop produces the 50% gray value from equal amount of RGB. To make sure I had as little color bias as possible I used 55K light sources and balance for the light source. When the image move to photoshop, I used the Eye Dropper to for any minor corrections for RGB, the Gray card was only off by two points on one color, before correction to RGB.
    It is odd that my meters with calibrations around 12.5 yielded middle gray, I suspect that these differences in statement of middle gray are still intended to be middle gray, but that the value come from different log scales with different number steps, or log bases.
    i just found this statement on Widipedia ( http://en.wikipedia.org/wiki/Middle_gray ) : "As early as 1903, middle gray was defined as the geometric mean intensity between a white and a black intensity that are in a ratio of 60:1.[5] That is equivalent to 12.9% of the white intensity"
    So I am wondering if people aren't making more of this than is 12-18% issue than is needed, half way between white and black is middle gray no matter how many steps you use to divide it in (as long as it is an odd number).
    The reason why it's useful for me is, I can look at a scene and if I can count on my exposure for a mid tone exposure, I can see the 18% card in my mind fairly well, I then can place exposure compensation while looking through the viewfinder, placing a tone where it should be fairly close, rather than shooting and then inspect the histogram. It allows me to get much better results if my time with the subject is time limited.
    thanx
     
  8. Vary true Steve,
    Ansel Adams was vary aware though of the K factor though even in the fifties. Adams also well supported placing exposure were you want for expression, as does any photographer in not too much time in their growth.
    The monitor issue is not relative i think, Photoshop produces the 50% gray value from equal amount of RGB. To make sure I had as little color bias as possible I used 55K light sources and balance for the light source. When the image move to photoshop, I used the Eye Dropper to for any minor corrections for RGB, the Gray card was only off by two points on one color, before correction to RGB.
    It is odd that my meters with calibrations around 12.5 yielded middle gray, I suspect that these differences in statement of middle gray are still intended to be middle gray, but that the value come from different log scales with different number steps, or log bases.
    i just found this statement on Widipedia ( http://en.wikipedia.org/wiki/Middle_gray ) : "As early as 1903, middle gray was defined as the geometric mean intensity between a white and a black intensity that are in a ratio of 60:1.[5] That is equivalent to 12.9% of the white intensity"
    So I am wondering if people aren't making more of this than is 12-18% issue than is needed, half way between white and black is middle gray no matter how many steps you use to divide it in (as long as it is an odd number).
    The reason why it's useful for me is, I can look at a scene and if I can count on my exposure for a mid tone exposure, I can see the 18% card in my mind fairly well, I then can place exposure compensation while looking through the viewfinder, placing a tone where it should be fairly close, rather than shooting and then inspect the histogram. It allows me to get much better results if my time with the subject is time limited.
    thanx
     
  9. Thanks John, I actually was thinking about that as a process on my own just yesterday, how some people say 11 stops and some say 7 stops is the dynamic range of the digital image. I know the answer for the larger value was in the RAW file data. So I did that in 1/3 stop increments.
     
  10. Randy, thanks for the Wikipedia quote. Now it all makes sense! Previously it had been one of those never-ending discussions like DPI on the digital darkroom forum (yawn).

    Other participants, if applicable, which gray square best matches your Kodak gray card?
     
  11. I don't know if anyone gets the irony here? Mid gray is mid gray is mid gray. The only reason I believe that the numbers come up different is if you are working in one environment (1.8 vs 2.2 gamma) and "assign" the other. Otherwise, in a color managed environment they will be the same. For instance, I work in a 1.8 gamma profile, on a screen calibrated at 2.2 gamma and print in both environments. Doesn't matter, that is what the profiles are for. But before the Lambda printers could read color profiles and with the Costco/quick print machines that still can't, if I don't convert to a 2.2 gamma space like sRGB, then my prints will be dark--otherwise they are fine.
    If mid gray were different on different machines, each would create a different photo density for output and then we would be in deep doo-doo!
     
  12. Well , I am just about to throw in the towel here. Sounds like the problem is that over the years of digital development, that a straight forward universal standard of how the eye sees luminance was abandoned and manufacturers went kind of willy nilly.
    Still mid gray within the digital camera you would think would be presented mid gray on the histogram on the camera. Other wise we don't know how the camera meter is performing, how much correction should be made, or if it should be made. We can examine highlight details and shadow details and base ball park exposures on that, but it seems that to reproduce a specific precise tone is rather luff.
    There are subtle differences just from one image to the next, not having moved lighting, subject of camera, and the range seems to float about + or - 1/4 stop in difference. Most of the time my camera says middle gray is 1/3 -1/2 stop lower than the histogram suggest it is.
    John I am reading an article that says we are in deep doo-doo. PDF download, about reproduction of art work troubles meeting museum quality standards. "Adopting ISO Standards for Museum Imaging"
    http://www.cdiny.com/ArticlesWhitePapers/ISO%20Standards%20for%20Museum%20Imaging_cdi_v1.0.pdf
     
  13. >> Well , I am just about to throw in the towel here. <<
    Hi Randy, it seems hard to get good information once you try to dig below a certain level. Let me try to steer you, fairly concisely, to the middle-grey thing.
    First, I don't know exactly what Photoshop's middle gray means. But, if you told me that you are using sRGB, and mean RGB pixel values of 127, 127, and 127, ok. sRGB is based on human perception (colorimetery), not exactly on measured reflection. But there's a way to calculate things backwards, at least within reason.
    It's much easier to go to Bruce Lindbloom's excellent web site and use one of his calculators. sRGB values 127, 127, 127 result in a CIEXYZ value for 'Y'= 21.5, which is basically equivalent to 21.5% neutral reflectance. If you start with sRGB values = 117 or 118, all across, this is roughly equivalent to .... 18% reflectance!
     
  14. deleted double post...
     
  15. That PDF ISO Standards for Museum Imaging seems really important. "Consider adopting the eciRGBv2 working space" instead of AdobeRGB, hmm.
    Bill C, I don't understand how you got 117 or 118 from Bruce Lindbloom's companding calculator. That would be darker than a gray card looks, even on a 1.8 gamma adjusted monitor.
     
  16. >> I don't understand how you got 117 or 118 from Bruce Lindbloom's companding calculator. <<
    Oh, sorry, not the companding calculator. You could use the "CIE Color Calculator", like so:
    - set your "RGB Model" (2nd box from bottom) = "sRGB"
    - go to the "RGB" row (3 boxes higher); at far right, check the "scale RGB" box
    - in the three empty RBG boxes, enter 118 (separately, in each box)
    - to run the calculation, click the RGB label/button. This should populate all the boxes.
    - note the top row, the XYZ data; the center box, the 'Y' value, is about 0.184, or 18 1/2, depending on whether the "scale XYZ" box is checked, or not. In the case of a neutral reflector, the 'Y' value should be equivalent to the reflectance, ~ 18 1/2%, in this case.
     
  17. Well it is always a little disruptive when while standing there viewing the world from your perspective and someone pulls the rug out from under you and you fall on the floor with a momentary heightened sense of awareness.
    using a suggestion from "John A" in hopes of re-centering my world, I shot a white towel in 1/2 stop increments over and below this 12% "middle gray value that lies off center of the histogram to the left" and found that I did indeed get about 2 1/2 stops of details without clipping, above or below this "middle gray value that lies off center of the histogram to the left". We got to come up with a better name for it though.
    Hmmmm
    Thank for all your input everyone.
     
  18. I believe this is the best discussion of this question that I've seen. I have been struggling with the details and I ALMOST follow the discussion. I wonder if one of you can just answer a simple question for me. I'm trying to confirm the calibration of my light meters (both are Sekonic). If the meter is accurately calibrated and I shoot an 18% gray card based on the meter's readings, should the resulting image be 127,127,127 in Photoshop?

    I know the question sounds simplistic, but inherent in the question is the entire discussion above. The implication from the question is that if the resulting image is not 127,127,127 then I need to adjust my light meter's calibration to get it to be so.

    Studying this question for some time, I've realized that there are still plenty of subjective factors that come into play, such as, once calibrated, do I point the dome at the light source, the camera or somewhere in between? That I can deal with, but not knowing precisely what is happening with the meter leaves me feeling somewhat limited.
    Thanks all.
    Bruce
     
  19. I am having some trouble following this thread almost 3 years later. I think the 18% number is hogwash. The right number is 12.9% for metering. Thom Hogan wrote this interesting essay about it, with several references to photo.net!
    To answer your question, I think the numbers are 117,117,117 approximately.
     

Share This Page