What constitutes a 'correct' histogram?

[george013]

I would have to STRONGLY disagree with this statement. A high key image is a properly exposed image that predominantly has light colors shot against a light coloured background. A low key image is a properly exposed image that predominantly has dark colors shot against a dark background. What you are describing is incorrect. Loss of detail at that scale is quite unacceptable in any image.

Massive overexposure is just that and does not result in a high key image. Massive underexposure does not result in a low key image.

May I add that a high key or low key image is more a subject choice than a shooting technique.

George

[Brev00]

There is also the case of silhouettes which can be created at the time or in post by clipping blacks intentionally. Who does not like a sharply edged black figure against a sunset scene? (That is a hypothetical question--no answer needed). So, I would just like to add that it is all about my intention, whether I clip blacks and/or whites, keep my histogram as shot, adjust the white and black points in post, or delete and start again. I do like to view the histogram when I take my image and when I adjust settings in post. I may allow clipping in my lcd as I find those tones are not always clipped in my raw file. Taking pics of flowers or folks wearing red shirts is a separate matter, as has been mentioned, and may require -ev and the use of the rgb histogram. Again, some clipping warnings may be louder in your camera than when viewed on your computer.

[xpatUSA]

Alan - the histogram tells you a lot of things, but there is no "right shape" to one. The histogram is what it is.

By the way, there are two types of histograms, there is a luminance histogram (I assume this is what you are looking at) as well as a set of histograms that show the distribution of each of the three colour channels.

Is there not a third type, known as "RGB" - the sum of the individual channel histograms:

https://www.cambridgeincolour.com/tu...istograms2.htm
.

[Manfred M]

Is there not a third type, known as "RGB" - the sum of the individual channel histograms:

https://www.cambridgeincolour.com/tu...istograms2.htm
.

All three channels is the Luminance histogram. I suspect there may be others, but my camera only provides two view - all three channels blended and a histogram of the individual R, B and B channels.

[dan marchant]

I would have to STRONGLY disagree with this statement. A high key image is a properly exposed image that predominantly has light colors shot against a light coloured background.

No it isn't. High key and low key are lighting techniques.... the tones of the scene are irrelevant. I can shoot a high key photograph of a black man standing against a dark wall.
https://en.wikipedia.org/wiki/High-key_lighting

[george013]

No it isn't. High key and low key are lighting techniques.... the tones of the scene are irrelevant. I can shoot a high key photograph of a black man standing against a dark wall.
https://en.wikipedia.org/wiki/High-key_lighting

The first sentence from your link

Not to be confused with High key.

George

[dan marchant]

The first sentence from your link

George

You are correct - I posted the wrong link.

The correct link is https://en.wikipedia.org/wiki/High_key which states that the image contains primarily light tones - not that the original scene had light tones. You achieve a high key image by overlighting the scene or adjusting your camera to "over expose" - not that there is actually such a thing as a correct exposure beyond whatever the photographer wants the exposure to be.

[dje]

Is there not a third type, known as "RGB" - the sum of the individual channel histograms:

https://www.cambridgeincolour.com/tu...istograms2.htm
.

Yes Ted I believe the RGB histogram referred to in the tutorial is one of the options found in Photoshop and FastStone Viewer (and several other pieces of software I would think). It's an unusual histogram in a way because it is, as you say, the sum of the individual colour histograms. My understanding is that it works like this :-

Say there are 20 pixels with R value of 100, 30 pixels with B value 100, and 40 pixels with G value of 100, then the histogram y value for an x value of 100 would be 20+30+40=90. The main point about this is that the 30 pixels with B value of 100 are a different set of pixels to those with an R value of 100. Similarly for the G values. So the histogram has lost any relationship to the individual pixels. It still will show highlight clipping though but with no information on which colour channels are clipping.

The luminance histogram is the only one that shows a display of how the "brightness" of the pixels is distributed.

Dave

[george013]

You are correct - I posted the wrong link.

The correct link is https://en.wikipedia.org/wiki/High_key which states that the image contains primarily light tones - not that the original scene had light tones. You achieve a high key image by overlighting the scene or adjusting your camera to "over expose" - not that there is actually such a thing as a correct exposure beyond whatever the photographer wants the exposure to be.

I wonder how you can create a high-key image of a black man against a black wall. The result won't be a black man against a black wall.

George

[xpatUSA]

Yes Ted I believe the RGB histogram referred to in the tutorial is one of the options found in Photoshop and FastStone Viewer (and several other pieces of software I would think). It's an unusual histogram in a way because it is, as you say, the sum of the individual colour histograms. My understanding is that it works like this :-

Say there are 20 pixels with R value of 100, 30 pixels with B value 100, and 40 pixels with G value of 100, then the histogram y value for an x value of 100 would be 20+30+40=90. The main point about this is that the 30 pixels with B value of 100 are a different set of pixels to those with an R value of 100. Similarly for the G values. So the histogram has lost any relationship to the individual pixels. It still will show highlight clipping though but with no information on which colour channels are clipping.

That's how I understood it too.

The luminance histogram is the only one that shows a display of how the "brightness" of the pixels is distributed.
Dave

Yes, which can lead to misunderstanding because of the weighting given to pixel values (approx. 0.2Red, 0.7Green, 0.1Blue). A little thinking reveals that a literally blown sky would show up toward the left of a luminosity histogram but somewhat darker greenery would show up toward the right. And attempting to ETTR in such a scene would not exactly be easy.

Looks like Manfred was talking about his own camera histograms, one being "all three channels blended" which I suspect refers to RGB, not luminosity. Could be wrong there, I don't own his camera models.

[edit] A little further investigation reveals that my Sigma image editor review histogram has a selection called "Intensity" which WikiPedia defines as the average of the three pixel values. Good luck with checking for color clipping with that! For example (255+255+6)/3 = 172 = apparently not clipped. Bloody useless!

Another web page mentions a histogram based on B, as in HSB (or V in HSV). If it's checking for clipping you're after, that would be your man because B (or V) is simply the highest value in a pixel. [/edit]

[Mike Buckley]

There are so many conflicting definitions of high-key and low-key images that I don't believe it is possible to arrive at a consensus definition. In the end, the label applied to this or that image doesn't matter; the only thing that matters is that we can consistently produce a certain look that we find appealing for whatever reasons.

[xpatUSA]

There are so many conflicting definitions . . .

. . . of high-key and low-key images that I don't believe it is possible to arrive at a consensus definition. In the end, the label applied to this or that image doesn't matter; the only thing that matters is that we can consistently produce a certain look that we find appealing for whatever reasons.

Yes, the first part of that comment applies to just about anything in photography, eh?

I agree with your conclusion, too, in spite of my many pedantic posts here.

I found this explanation to be quite reasonable:

http://www.nobadfoto.com/histogram-2.html

CiC doesn't do too badly either:

https://www.cambridgeincolour.com/tu...istograms1.htm (scroll halfway down).
.

[Manfred M]

Looks like Manfred was talking about his own camera histograms, one being "all three channels blended" which I suspect refers to RGB, not luminosity. Could be wrong there, I don't own his camera models.

Is it "true" luminosity? I suspect the answer is "no", as the data is drawn from a jpeg that is some flavour or RGB. True luminosity would have to come from a LAB colour space as would have to show the L channel. At a high level, it would have to be looked at as an approximation of luminosity, but other than for the pedantic crowd, it would be close enough.

Unfortunately, camera manuals don't go into any details here (we wouldn't want to enlighten the users).

[Manfred M]

No it isn't. High key and low key are lighting techniques.... the tones of the scene are irrelevant. I can shoot a high key photograph of a black man standing against a dark wall.
https://en.wikipedia.org/wiki/High-key_lighting

I'm going to have to go back and argue with my Photographic Lighting prof (and the college), as those were the definitions we were given during the Studio Lighting course.

I also find it interesting that the one thing the Wikipedia article concentrates on is that the technique was used in order to get correct exposure because the image capture tubes in the 50's and 60's, as the article says it so well "did not deal well with high contrast ratios", which is a very convoluted way of saying that they had a very limited dynamic range. Then, as now, the key criteria was to have a well lit scene that could be captured by the equipment from the black values through to white values without clipping at either end. Some things haven't changed; proper exposure was (and still is) key to getting a good image.

I also find it interesting that both images that are show in the examples in the article meet the definition that I used to describe high key images - predominantly light subjects shot against a light background.

Everything I see in the article suggests that they were trying to avoid blown out highlights and blown out shadow detail. An image with a lot of either is not a technically good image, although I don't see an issue with a very limited amount of clipping of highlights and shadows, as long as the image works. I've had the concept of proper exposure pounded into my head over far to many years.

[Inkanyezi]

Whatabout a spanner in the works here?

Considering the technical properties of our medium, the photon well, a bucket that can be filled, there is a possibility of the bucket flowing over, There is however also the possibility of an empty well, and there is a huge difference in how the two conditions are achieved. In an electrical engineering sense, the left side of a histogram by definition cannot be clipped. The photon well might be empty, which in practice is a total blackness for that colour, and if all three colours are empty, we have a total black. To me, that does not mean "clipping", only that no light was captured.

The [b]right end of the histogram is quite another thing. There's where clipping can occur. A well that flows over indicates that there was light there, which could not be registered by the well, because it flowed over. This can happen to any individual colour, as well as two of them or all three.

Using the term "clipping" for the dark end of the histogram is highly confusing and in my opinion it should not be used; to me, it is dead wrong.

The photographer is free to interpret the scene. If I want part of it vanish in total darkness, I am free to do so. This will certainly, but not necessarily, be the case in many low key images. OTOH, if I wish parts of the image be lost in total brightness, I have the same freedom to do so as well, which in itself does not indicate that the image is high key. In many cases such an approach could destroy the intentions of the photographer, who might have preferred a gradual approach until only a infinitesimal part of the image reaches total brightness, which indicates that no clipping takes place. In my opinion, clipping occurs only at the right side of the histogram, as by definition, it cannot occur at the left side. An empty bucket is just empty, while an overflowing one spills over. The high key approach is a bit more difficult, as far as colour is concerned. All three colours must reach the full state simultaneously, in order to produce pure white. In my opinion, high key images should ideally not be washed out (YMMV).

The term "clipping" is an electronic engineering term, which was not present in older theory with physicalic registering of images and subsequent chemical processing. There is no mention of clipping in Ansel Adams' The Negative or The Print. Zone zero is just zero, and zone X is just zone X. In digital imaging, a large field of clipped pixels falls into zone X, and a huge block of empty wells will fall into zone 0. Ansel Adams was aware of the negative able to register brighter values than zone X, but difficulty to render them in the image with normal processing. In the past we used tone separation to accomplish that, and in digital photography the approach is tone mapping, which enhances local contrast, while containing all luminosities within the global tonal range of the display medium. To do that, those bright parts of the scene must not be clipped.

[pnodrog]

A definition for clipping as applied to photography "An image where the intensity in a certain area falls outside the minimum and maximum intensity which can be represented". It is not unreasonable to use this term if image detail is lost either in the shadows or highlights. When looking at a histogram and you see a significant level at either end, it may correctly indicate that for the exposure given that image information extends beyond the sensitivity and/or dynamic range of the sensor. Whether the information lost is deemed significance will depend on the photographer. (or critical viewers...)

[Inkanyezi]

A definition for clipping as applied to photography "An image where the intensity in a certain area falls outside the minimum and maximum intensity which can be represented". It is not unreasonable to use this term if image detail is lost either in the shadows or highlights. When looking at a histogram and you see a significant level at either end, it may correctly indicate that for the exposure given significant image information could have extended beyond the sensitivity and/or dynamic range of the sensor.

So, I bolded the important part there. To me - note that this is not a general "to photography" assertion - there is no clipping other than the signal clipping that occurs where there is a signal that is beyond the scope of registration. At the dark end of the histogram, there is no signal, or at least no detectable signal. To me, that is a vast difference, and the notion of "clipping" at the left end is confusing and to the novice, it obscures more than it explains.

[george013]

So, I bolded the important part there. To me - note that this is not a general "to photography" assertion - there is no clipping other than the signal clipping that occurs where there is a signal that is beyond the scope of registration. At the dark end of the histogram, there is no signal, or at least no detectable signal. To me, that is a vast difference, and the notion of "clipping" at the left end is confusing and to the novice, it obscures more than it explains.

What I learned is that on the left hand the noise signal determines the minimum.

George

[dje]

I think where the term "black clipping" comes in is where the black point has been lifted with post processing software. Under these circumstances, there is a range of values close to black that are re-mapped to black (zero). This is somewhat similar to white clipping in that there is a range of values all mapped to white. Both ends can exhibit spikes in the histogram. That is not to say that the effects are the same though.

These days, it seems that terminology can mean whatever you want it to mean!

Dave

[pnodrog]

Urban it is a bit unusual but I don't think we are going to agree...

But simply put if the histogram show significantly high values in the deepest shadow area it indicates the exposure may need to be increased to capture any lost shadow detail. If the histogram shows significantly high values in the highest highlight area it indicates that the exposure may need to be decreased to avoid loss of highlights and if the histogram has significantly high values at both extremes it indicates the scene has a high dynamic range and the photographer needs to choose whether to lose shadow detail, blow out the highlights or use an HDR technique.