Reproduction of Colour – Gamma for slides

[guyburns]

This may not be the best place for my questions, but I couldn't find a "technical" section in these forums. If there is a better place, please let me know.

I'm working my way through The Reproduction of Colour by Hunt, and when I come across something I can't understand, I intend to ask the experts. On page 57, Hunt is describing why slides need to have increased gamma when projected in dark surrounds. The description is along the lines of:

The effect of a dark surround [the dark room in which the slide is being projected] is to make the image appear lighter… To produce an image that appears correct requires an increase in gamma to about 1.5… In order to achieve a gamma of 1.5 when projected, it is necessary to have even higher gammas at the capture stage in order to overcome the effects of flare light in the camera and in the viewing situation…

The accompanying figure is attached below. The horizontal axis is scene brightness (increasing to the right in log units), and the vertical axis is film density, also in log units, with the film getting blacker the higher you go.

What I don't understand is why the "ideal system" in the graph trends vertically. i.e. there comes a point at a scene luminance of about -1.4 where all parts of the scene below that brightness will be recorded on film as effectively black. Hunt reckons that is the ideal. Why would it ever be ideal to toss out shadow information?

[crames]

What I don't understand is why the "ideal system" in the graph trends vertically. i.e. there comes a point at a scene luminance of about -1.4 where all parts of the scene below that brightness will be recorded on film as effectively black. Hunt reckons that is the ideal. Why would it ever be ideal to toss out shadow information?

Flare ultimately limits the darkest tone that can be reproduced. The shadow information that is being tossed out is shadow information that is darker than the darkest tone that can be reproduced by a system with the various sources of flare described in the text.

"Reproduced" being defined by Hunt as: the reproduction having the same relative luminances as the original scene.

[guyburns]

First to answer -- my old mate Crames from the Kodachrome thread! I thought if anyone could answer questions from Hunt it would be you. Stick around, because I'm only up to page 57 and there are another 650 to go. I've been through the book once, and am now on a second, very detailed pass.

Your answer seems reasonable. I half suspected that Hunt was cutting off the blacks because there would be no point in having them (for the particular situation he described) if they are washed out by other light sources, but I was not convinced that such an approach would be ideal so I assumed I was missing something. But I think you are right, which means Hunt is wrong in ascribing such a curve as ideal. Rather silly of him to describe a slide with a log-luminance capture range of 1.4 as ideal because slides have other uses than being projected. No one would choose to implement his "ideal" slide characteristic even if it was possible. He should have used a word other than "ideal".

[crames]

First to answer -- my old mate Crames from the Kodachrome thread!

Yes, the Kodachrome threads - after those I swore to myself that I wouldn't jump into any more threads!

I thought if anyone could answer questions from Hunt it would be you. Stick around, because I'm only up to page 57 and there are another 650 to go. I've been through the book once, and am now on a second, very detailed pass.

Pretty much what I have done, then re-reading some chapters of special interest multiple times. There's an incredible amount of information in that book, and I certainly don't understand more than a small fraction of it.

Your answer seems reasonable. I half suspected that Hunt was cutting off the blacks because there would be no point in having them (for the particular situation he described) if they are washed out by other light sources, but I was not convinced that such an approach would be ideal so I assumed I was missing something. But I think you are right, which means Hunt is wrong in ascribing such a curve as ideal. Rather silly of him to describe a slide with a log-luminance capture range of 1.4 as ideal because slides have other uses than being projected.

Now, don't you go dissin' my man Hunt!

I think he's just being realistic in showing what is achievable if the goal, as in most of the examples in that chapter, is to approximate the appearance of the original scene. I suppose that if slides are to be used other than for projection, it would be better if they had a much smaller gamma than they do.

No one would choose to implement his "ideal" slide characteristic even if it was possible. He should have used a word other than "ideal".

Well, he does show an "actual film" curve that closely adheres to the the ideal curve until veering off after 1.4, so at least one commercially successful amateur slide film has attempted the "ideal."

[guyburns]

I have a few graphs and some maths (see attachments) to see if I have a good grasp of these characteristic curves.

At the top of page 61, Hunt talks about how light from the screen which is reflected by the room back to the screen (viewing flare), can alter the appearance of the image. In particular, a darker image will throw less light off the walls:

hence the viewing flare will decrease somewhat… and the effective gamma will rise at high densities.

Well, I don't always believe your man Hunt, and since he previously hadn't explained why the gamma rises at high densities if there is less viewing flare, I wanted to prove it myself (attachments). Leaving the maths to one side for the moment, I assume that if there is less viewing flare, then darker parts of the image which would have previously been washed out, will now show visible detail, and in some way the gamma is thereby increased. On the other hand, because the surround is now darker, that should mean a slight decrease in effective gamma (the darker the surround, the more gamma has to be increased to compensate). I assume the former effect overwhelms the latter.

In the attachments, I have drawn a graph similar to Fig 6.12, then tried to work out what would be the effect of adding viewing flare of 5% and 2%. If my maths is correct, the effect is just as Hunt says it would be: in going from 5% to 2% (a reduction in viewing flare) the on-screen gamma increases.

I'll never doubt your man again.

[crames]

The calculations look good to me, and that's an interesting point about flare possibly contributing to the surround. A change of a few percent in the surround luminance could have an effect, although maybe not enough to change the category of the surround from dim to dark.

A little off topic, but now that you're thinking about the effect of the surround on viewing contrast, what do you think about using a monitor in a dark surround to edit an image destined for display in an average surround?

[guyburns]

I could make a few guesses about the appearance of monitor images with different surrounds, but I'll refrain until I feel confident about answering -- which won't be until I finish Hunt. If you can remember my Kodachrome post, I'm actually on a search for a more complicated translation: mimicking the look of projected Kodachrome slides on a digital cinema projector, after scanning, colour correction and gamma correction.

[crames]

If you can remember my Kodachrome post, I'm actually on a search for a more complicated translation: mimicking the look of projected Kodachrome slides on a digital cinema projector, after scanning, colour correction and gamma correction.

Yes, it can get complicated. It could be as simple as applying just a chromatic adapatation transform before projection - depends on what you goals are.

[guyburns]

A little off topic, but now that you're thinking about the effect of the surround on viewing contrast, what do you think about using a monitor in a dark surround to edit an image destined for display in an average surround?

I'm back again Crames. Just finished Hunt – what a battle. As for your question, I'll assume that the theory applied by Hunt to projected images applies to monitors. Given that an image should have a gamma of 1.5 when viewed in a dark surround, and 1.25 when viewed in a dim surround (your "average" surround?), that means that an image you are happy with when viewed on a monitor in dark surrounds should have a gamma correction applied of 1.25/1.5 = ~0.8. I assume you could apply that in Photoshop via a Levels Adjustment layer by setting the middle-slider to 0.8.

I'll back this up by what Hunt says on page 439, but in a different situation (this quote may have been slightly edited):

The signals from colour television cameras are nominally gamma-corrected to the power of 1/2.2 (0.45); this gamma, when combined with a receiver gamma of 2.8, gives a system gamma of about 1.25, as required for dim surrounds. If pictures derived from film are to look the same as those derived from television cameras… the signals obtained from the film must also possess a gamma of 1/2.2. Most film is made for projection in darkened rooms, and for this purpose it requires an effective gamma on the screen of 1.5… gamma correction amounting to 0.45/1.5, which is equal to 0.3, must be introduced [(1.5 x 0.3 = 0.45)].

[crames]

I'm back again Crames. Just finished Hunt – what a battle. As for your question, I'll assume that the theory applied by Hunt to projected images applies to monitors. Given that an image should have a gamma of 1.5 when viewed in a dark surround, and 1.25 when viewed in a dim surround (your "average" surround?), that means that an image you are happy with when viewed on a monitor in dark surrounds should have a gamma correction applied of 1.25/1.5 = ~0.8. I assume you could apply that in Photoshop via a Levels Adjustment layer by setting the middle-slider to 0.8.

Yes, if the surrounds are different, there has to be a gamma correction, otherwise the contrast will be different. I think this is the biggest cause of mismatches between prints and monitors. Discussions focus on adjusting the absolute luminance of the monitor, when what matters is the ratio of the monitor luminance to the surround luminance.

Prints are normally viewed in "average surround" conditions (surround luminance at least 20% of the image white). Unless the monitor is also viewed in an average surround, it will show weaker contrast than the print. An image that looks good on the display in a dim (<20%) or dark (0%) surround will have too much contrast when printed and viewed in an average surround. To compensate for a dim surround, apply .8 gamma before printing (1.25 on the Levels middle slider, which is 1/gamma). For a dark surround, apply a .67 gamma before printing (1.5 on the Levels middle slider).

Sorry, off-topic but hard not to think about these kinds things after reading Hunt!