18% grey card image

[john_oconnor3]

I'm not sure where to actually post this query but here goes. If I expose a

grey card in good light using a 400d and 50mm/1.8 and then download into PSE5 I

get a greyscale reading for the narrow peak of 103 instead of the expected

128. I have no way of explaining this and I wonder if those more

knowledgeable than I can help. I dont feel my "normal" pictures are

underexposed and as I normaally shoot in RAW I can easily make adjustments if

they were. Has anybody found a similar situation and can you explain it! Best

wishes John

[peter_rowe]

John,

I believe the 18% gray card may be a little off in that many manufacturers use 15% which would be a little brighter. There is some complexity behind this that I am sure others will jump in with.

[curt wiler]

If a gray card is labeled 18%, then that is what it is, aside from errors due to surface sheen, etc. If you are using the camera's meter to determine the exposure, then you should be aware that the standard for meter calibration is closer to 13%, so there should be an offset of about 1/2-EV.

[rodeo_joe1]

Firstly John: You're forgetting the gamma introduced by whatever colour space your camera is set to. For the sRGB space the gamma is 2.2 which means that 18% density should have a pixel value of 255*((18/100)^1/2.2) or ~ 117.

 

With the more sensible and printable Apple gamma of 1.8, it should be 255*((18/100)^1/1.8) = 98.

 

I don't know why you think it should be 128 since even with a gamma of 1, it would be 255*0.18 = 46.

 

Incidentally, a reflective card value of 18% is a "bastard" value that doesn't fit in with any sensible system of sensitometry. The ISO speed system is based on a mid grey which is 4 stops up from the toe value of conventional film, and 3 stops down from maximum exposure. This should give a reflectance of 12.5% by anybody's reckoning.

 

A more logical system for digital camera use would be to take a white card reading and add 3 stops. This is because digital cameras have very little overexposure latitude, and also because a white card will give you a reading that's accurate to within 5% unless it's visibly dirty. A grey card can fade or get dirty to a tolerance of several percent before you notice the difference, and 18% plus or minus 5% gives a lot greater error than 100% - 5%.

 

(5% in 18% = 27% error, whereas 5% in 100% = 5% error).

 

Plus, a simple doubled-up piece of copier paper will give you a near enough perfect lambertian reflectance of around 98%; costs next-to-nothing and can be discarded and replaced as soon as there's a hint of dirt on it.

[beauh44]

In addition, according to Lee Varis, the author of "Skin" (an excellent book on getting accurate skin tones with digital cameras and Photoshop) the Kodak gray card isn't "spectrally neutral" and instead he recommends a "Robin Myer Digital Gray Card" from http://www.rmimaging.com I have no idea about its RGB values as far as 18% etc. goes though.

[bobatkins]

The card probably isn't 18% and even if it was the greyscale value wouldn't be 128. As "Rodeo Joe" points out, gamma is involved.

[mendel_leisk]

Thanks for the formula info, Rodeo Bob. Just reverse engineering, I see "^" represents "to the power". New one for me.

[garydem]

personally, i use a white card not a grey one.

 

also for your test you said "good light". what you really need to know is the amount of lumens falling on the card during your test shot. and then compare that number to the situation that the card was made for at the factory. they had to use a test chamber of some kind with some amount of lumens falling on their card. the problem is you have no way of knowing what their test situation was.

[bobatkins]

Illumination level won't affect reflectivity. If it's 18% under bright light, it's 18% under dim light too, assuming both the bright and dim light have the same spectral distribution.

 

If the card isn't perfectly neutral (i.e. reflectivity varies slightly with wavelength), which is often true, then you might get slightly different reflectivity for, say, tungsten, fluorescent and daylight. So it might be 17% under one type of lighting and 19% under another.

[blumesan]

@ Rodeo Joe

 

Thanks for a very cogent post on this subject. I believe you have made one typo in your suggestion to use a white card reading:

 

"A more logical system for digital camera use would be to take a white card reading and add 3 stops."

 

Did you perhaps mean to say: "SUBTRACT 3 stops"? Or perhaps I am interpreting it incorrectly.

[colinsouthern]

"Did you perhaps mean to say: "SUBTRACT 3 stops"? Or perhaps I am interpreting it incorrectly."

 

If you point a camera at a white card it'll expose it as a grey card - so you need to ADD 3 stops of light to shift it up to where it needs to be.