Human Eye Response

GA
Posted By
George_Austin
Nov 7, 2003
Views
856
Replies
18
Status
Closed
It seems to be deeply rooted in photographic lore that a card reflecting 18% of white light is perceived as 50% gray. For this to be true, the power to which the (normalized)measured light must be raised to give the perceived light would be 1/2.5. Yet, in the literature (specifically at www.lutron.com), the perceived power is given as the square root of the measured power (1/2). The latter gives a perceived gray level of 50% for a measured power of 25%. Which relation is authentic?

MacBook Pro 16” Mockups 🔥

– in 4 materials (clay versions included)

– 12 scenes

– 48 MacBook Pro 16″ mockups

– 6000 x 4500 px

L
Lundberg02
Nov 7, 2003
Now you’ve gone and done it, George
JS
John_Slate
Nov 7, 2003
More importantly, is 50% gray one half of white, or one half of black?

Maybe 50% gray is like a wormhole between additive and subtractive, dude.

Whoa!
GA
George_Austin
Nov 7, 2003
Lundberg02

"…Now you’ve gone and done it, George …"

Oh oh! I assume you mean this is an old argument raised too many times without resolution.

I did do a forum search but not an exhaustive one. I searched for "eye response" (1) with quotes and got none, and (2) without quotes and got too many to count and gave up when the first dozen proved way off the mark.
JS
john_schutkeker
Nov 8, 2003
By "Now you’ve done it," he means that you’ve just found yourself a nice dissertation project. I hope you live in a town that has a lot of university libraries.
GH
Gernot_Hoffmann
Nov 8, 2003
George,

Y=0.18 CIE XYZ luminance is converted to CIELab luminance by L*=116*(0.18)^0.333333 – 16 = 49.5 (nearly 50%).

Some people may consider this as a "medium gray" , but the appearance depends much on adaptation. For me it looks on a white background too dark .

Best regards –Gernot Hoffmann
CC
Chris_Cox
Nov 9, 2003
George – no argument. 18% reflectance looks to the human eye like 50% gray (assuming normal viewing conditions, "average" surround, etc.).

It can vary a bit due to viewing conditions (most notably in VERY low light – like from a single candle).
GA
George_Austin
Nov 9, 2003
Chris,

If 18% intensity is seen by the eye as 50% gray, how can the gray value 128/128/128, which is 50% intense also be seen as mid-gray? And why isn’t 46/46/46, which is 18% as intense as 255/255/255, seen as 50% gray?

Whether 128/128/128 "looks" mid-gray is debatable, but there is no question whatsoever that 46/46/46 is far too dark to be mistaken for mid-gray.

I am tempted to suggest that light "intensity" is the square (power 2)or some other power function near the square (2.5?) of the tonal value. At 2.5, a 128 tone would then be 18% of full intensity (.5 raised to the power 2.5 = .18). But it is easier to believe that intensity is the square of the tonal value and that the eye’s response accounts for the observed 2.5 effective power exponent.

George
GA
George_Austin
Nov 9, 2003
Gernot,

You’ve probably got the answer wrapped up in your reply—if only I could understand it 🙂 !

George
GH
Gernot_Hoffmann
Nov 9, 2003
George,

these are the magic numbers:

R=G=B=Y=0.18 ————-linear light space

R’=G’=B’=Y^(1/gamma) —-the numbers in the file

L*=1.16*Y^(1/3)-0.16 =0.5

Can be scaled by 255 for RGB and by 100 for L*.

For me, an L*=50 patch in the neighbourhood of other patches, adapted to a light background, looks too dark.
Some examples are here, especially a measurable collection of CIELab gray patches for the check of monitor calibration by instruments:
<http://www.fho-emden.de/~hoffmann/caltutor270900.pdf>

Best regards –Gernot Hoffmann
CC
Chris_Cox
Nov 11, 2003
George – you’re comparing apples and oranges.

18% reflectance is dealing with linear light (gamma 1.0).

That corresponds to something close to value 128 if you’re using a gamma encoding value close to 2.5.

Images are gamma encoded to make them closer to what we perceive (and thus maximize the usage of the available bits).
GA
George_Austin
Nov 11, 2003
Mac

That is a terrific reference. Some of the optical illusions are mind-boggling. It doesn’t address my very focused question but it is nevertheless pertinent. Thanks.

George
GA
George_Austin
Nov 11, 2003
"…Images are gamma encoded to make them closer to what we perceive …"

I dunno, Chris. Seems to me that images are gamma-altered to counteract the warping introduced by the system’s non-linear response. It’s purpose is to restore the light distribution of the original scene. Whatever the eye/brain does in perceiving/interpreting it, you/ve got to start with an input uncorrupted by the device through which it passes. The eye/brain will then do its thing on a true input.

George
GH
Gernot_Hoffmann
Nov 11, 2003
George,

Yes, images are pre-distorted by RGB´-Number=SceneLuminance^(1/2.2) in order to show them on a monitor which has the equivalent distortion VisibleLuminance=MeasurableLuminance=RGB´-Number^2.2 .
(Luminances and numbers are normalized for the range 0…1).

"…Images are gamma encoded to make them closer to what we perceive …" Totally nonsense. Mr.Poynton´s Fairy Tales.

I´m pre-distorting artificial images, show them on a calibrated monitor and MEASURE the luminances by instruments. Result are (in limits) the ORIGINAL luminances.

Best regards –Gernot Hoffmann
MS
Michel_Sylvestre
Nov 11, 2003
Isn’t there some relationship with this and the way light meters (in or off camera) are calibrated?
CC
Chris_Cox
Nov 11, 2003
George – see <http://chriscox.org/gamma/>

The display gamma has nothing to do with why images are gamma encoded.

And despite Mr. Hoffmann’s protests, this is the truth.
(otherwise you need more bandwidth/bits to represent the same image quality)

The correction done by Photoshop to match the image to the display — THAT is intended to restore the scene to it’s intended luminance.
GA
George_Austin
Nov 13, 2003
Chris,

The link to your Gamma dissertation is much appreciated. I’ll need time to absorb it. But after a quick scan I am troubled by the following excerpt:

"…Human lightness sensitivity has a gamma somewhere between 2.0 and 3.0 depending on the viewing conditions. Typical CRT gamma is between 2.0 and 2.7…These numbers are not designed to be close, but it just worked out that way…"

As I see it, the monitor and the eye, rather than being "similar", could hardly be further apart in their responses to input intensity. The monitor becomes increasingly sensitive as input increases, whereas the eye becomes inreasingly INsensitive as input increases. Their "gammas" are very nearly the INVERSES of each other.
CC
Chris_Cox
Nov 13, 2003
Hmm, yes – I could improve the wording there.

(gamma gets used as 1/gamma depending on the context – so it does get confusing sometimes)

MacBook Pro 16” Mockups 🔥

– in 4 materials (clay versions included)

– 12 scenes

– 48 MacBook Pro 16″ mockups

– 6000 x 4500 px

Related Discussion Topics

Nice and short text about related topics in discussion sections