Histogram: Luminosity and Color

[DanK]

Quote Originally Posted by DanK View Post
and you are being gratuitously disparaging because...?
Because what is quoted below is misleading. You mention a "peak" and you state that having that peak near the right is better than having it near the left.

First, that is no reason to be gratuitously nasty in your tone.

Second, you are right, I simplified, writing as if the distribution were unimodal and substantially nonuniform, which it is often is. So, substitute "mass of the distribution" for "peak of the distribution," and all should be fine.

Almost everyone on this site works to be constructive and pleasant.

[Inkanyezi]

Almost everyone on this site works to be constructive and pleasant.

Almost everyone, and you might allow a bit of leeway for a foreigner whose language is not the same as yours.

And when we're at it, maybe a bit of more explanation can be done about ETTR.

ETTR was an idea developed around early sensors that had a rather high noise threshold and narrow dynamic range.

Under those conditions, noise is often rather evident in the image, and it is more difficult to keep exposure within bounds, often impossible. Then, if noise is one determinant, keeping the histogram as far as possible to the right is rational, but pushing it too far right will cause clipping. So ETTR was the idea to increase exposure if possible, just as much as not to cause clipping. This was supposed to be optimal exposure, and it has become an often repeated mantra.

As can be seen in various posts in this discussion, the mantra of exposing to the right is not always productive in order to get good images. ETTR places the shadow areas higher when the subject range falls within the latidude of the sensor, thus suppressing noise by increasing signal to noise ratio. There are caveats, as seen in the discussion, most of them relating to the art of creating images.

So, if you bother about noise in dark areas of the image, increasing exposure if possible can be a good idea. But mostly, as noise in dark areas is not crucial to the impact of the image, reducing exposure and raising ISO in order to get shorter exposure time or a smaller aperture, more suitable for your subject, can be a better idea. Then, when your dynamic range is shrinking, you will perhaps want to expose to the right.

[Dave Humphries]

All,

It would undoubtedly help Maria if I showed the histograms for my two image examples.



As Urban says, the histogram of the swimmers shows two peaks, one on each extreme, because the scene is very 'non standard' in terms of the general advice about histograms.

The swan is more 'standard', but still shows two peaks, the one on the left, being lower luminance values, represents the water, while the one near the right end represents the white of the swan. It was important, when taking the image, that I did not over expose the white and lose all detail in the feathers. If I had over exposed, that could look quite like the right hand side of the swimmers histogram.

For anyone wanting to understand some of the technicalities related in posts above ...

However, to keep noise to a minimum, I would not really want to expose the swan any less than this, because that would make the white feathers grey - at least until I raised the exposure in Post Processing - but that would also have made the noise more obvious.

If the scene did not contain any white though, as also mentioned earlier, you would not want to expose too far to the right because it would not look correct and we would have to then bring the exposure down in Post Processing. It could be argued this would also reduce the noise, but I don't do this.

Someone mentioned that the "histogram is not scaled", what this means is that the vertical scale, or height, of the graph is not a standard numerical value (between different images). You can see this as I have placed my mouse cursor on the histograms at the highest counts of pixels (vertical), which is 29,268 on the swimmers and 14.423 on the swam and these occur at brightness values of 254 and 84 respectively (across the graph).

In other words; the vertical scale on one histogram is twice the other, but this doesn't matter, it is the shape of the curve that matters.

HTH,

[MariaMaria]

Thanks everyone for all your posts, suggestions, and feedback. Dave, It does help to see this histograms, thanks for those!

I must admit that I have been kind of lost with a majority of the "tech-talk" as you might say but I do appreciate everyone's feedback. Again, I am not an expert and at this point in my photo journey am just looking for general guidelines, rules-of-thumb, etc., to help me along. I plan to learn more of the tech stuff a little at a time so as to not become overwhelmed

Dan, I understood your ETTR suggestion as a "rule-of-thumb", that if my overall tonal range was limited, that I might want to ETTR rather than to the left to keep the signal-to-noise ratio higher (I found a tutorial about that on CIC ) It seems like ETTR is not something that would be appropriate for every single image that I shoot but might be helpful when the tonal range is limited. Is that an accurate understanding? If not be sure to let me know Dan!

However, to keep noise to a minimum, I would not really want to expose the swan any less than this, because that would make the white feathers grey - at least until I raised the exposure in Post Processing - but that would also have made the noise more obvious.

If the scene did not contain any white though, as also mentioned earlier, you would not want to expose too far to the right because it would not look correct and we would have to then bring the exposure down in Post Processing. It could be argued this would also reduce the noise, but I don't do this.

Dave, so in your above example, I interpret this as you exposed the Swan to the right which kept the feathers from being grey and the noise to a minimum; however, if you were shooting a scene with less white then ETTR might not be the best option?

So altogether from what I've read (and been able to understand) is that ETTR is not a one-size-fits all, but that it could be beneficial in scenes with a higher white content? I am going to have to read a tutorial just to understand this thread
However, the histogram can be used as a tool to monitor the frequency of my highlights/shadows at each tone and the RGB histogram will show me which channel may be blown/underexposed or if I may want to pay attention to that channel in PP? I hope that's close...

[Mike Buckley]

Maria,

It's clear from your most recent post that you're getting everything just fine, including the importance of ignoring the highly technical, highly nuanced stuff. The one thing that I would add is that ETTR will almost always be your better choice, especially at this stage of learning how to apply the information displayed in the histogram to the image. Another way of putting it is that if you err, err by ETTR. You'll learn over time when ETTR is not ideal.

[ajohnw]

Sorry Mike but I feel this area needs a bit more.

Something you may find odd about things like white swans or even flowers is that it can be better to under expose a little and brighten up in PP. That definitely leaves room for contrast type adjustments to enhance shape and say feather detail plus further brightening up if needed. The same sort of thing applies to any bright especially pale colours on subjects that don't have much inherent contrast.

The other problem of course is that not all cameras show a histogram in the view finder to allow it to be used to set an exposure so the ones in the preview have to be used. The histograms there can be used to judge any exposure increases that might be applied - if they have sufficient detail. Often they are just another way of showing clipping but blinkies and the overall appearance of the image is a better guide.

The other problem when cameras do have a histogram in the viewfinder is that they are sampling the image in places and not all over. Cameras are a lot better than they used to be but they can still miss things. So once again the preview has the final word.

The important thing about the bright end of things really is that any that are needed in the final shot must not be clipped because if they are there is no way of getting them back short of faking them by painting by hand.

PP Histograms are actual of what is in the shot and a very useful method of indicating what is going on when shots are adjusted.

John
-

[Mike Buckley]

Sorry Mike but I feel this area needs a bit more.-

Sorry, John, but nothing in your post contradicts my post. I wrote that "ETTR will almost always be your better choice, especially at this stage of [Maria] learning how to apply the information displayed in the histogram to the image." In my opinion, you and all the other folks who are very steeped in the nuances of ETTR and the relatively few situations when using it is not appropriate, are overlooking the context of the OP. While your posts are appropriate for some of the others following the thread, the most important posts in the thread in my opinion are those that respond directly to the concerns, experience and context of the OP.

[MariaMaria]

Maria,

It's clear from your most recent post that you're getting everything just fine, including the importance of ignoring the highly technical, highly nuanced stuff. The one thing that I would add is that ETTR will almost always be your better choice, especially at this stage of learning how to apply the information displayed in the histogram to the image. Another way of putting it is that if you err, err by ETTR. You'll learn over time when ETTR is not ideal.

Thanks Mike, this is reassuring

[Colin Southern]

The one thing that I would add is that ETTR will almost always be your better choice

Sorry, but I completely disagree.

ETTR carries risks as well as rewards, and with modern sensors, it's a risk that just doesn't need to be taken. Modern sensors are already capable of capturing far more dynamic range than we're capable or printing of displaying - so to deliberately bias the exposure upwards, risking capturing data in a sensor's non-linear region, only to have to bias it downwards in post-processing, risking issues getting the gamma right, just doesn't make any sense to me (unless it's a high dynamic range scene that requires it).

[Mike Buckley]

Sorry, but I completely disagree.

ETTR carries risks as well as rewards, and with modern sensors, it's a risk that just doesn't need to be taken. Modern sensors are already capable of capturing far more dynamic range than we're capable or printing of displaying - so to deliberately bias the exposure upwards, risking capturing data in a sensor's non-linear region, only to have to bias it downwards in post-processing, risking issues getting the gamma right, just doesn't make any sense to me (unless it's a high dynamic range scene that requires it).

I really do think you and others with experience that is similar to yours have understandably forgotten what the relative novice can absorb. First, if ETTR will not almost always be the better choice, then you should advise people to almost always NOT expose to the right. Second, you quoted me out of context, by eliminating my caution that my advice was offered in the context of the OP's current stage of the learning curve. Third, you ignored my recommendation that if one is to err, it would be better to err on the side of ETTR. I make that recommendation because my anecdotal experience is that far more photos are presented here under exposed than over exposed. Fourth, bringing the context of getting the gamma right is so far beyond Maria's current understanding that it's unhelpful to her.

Otherwise, your post was dead-on.

[Colin Southern]

I really do think you and others with experience that is similar to yours have understandably forgotten what the relative novice can absorb.

Just the opposite. If someone is just learning then to me it makes the most sense to NOT complicate things by suggesting that they start by manually over-riding camera metering at the time of exposure and then manually biasing it back the other way in post processing.

First, if ETTR will not almost always be the better choice, then you should advise people to almost always NOT expose to the right.

Yes. I'd advise them to not deliberately start biasing exposures - with all the risks that that entails - for the sole purpose of reducing noise that probably wouldn't have been an issue anyway.

Second, you quoted me out of context, by eliminating my caution that my advice was offered in the context of the OP's current stage of the learning curve.

Sorry, but I just think you offered bad advice (with or without the caution). I just don't see the logic in recommending people over-expose images and then have to fix them later if there's no real-world advantage in doing so. Or in other words, "no need to err on the side of over-exposure). Most cameras can capture 12 stops at base ISO, and most reflective scenes are only using 6 or 7 at most). The Camera isn't the limitation.

Third, you ignored my recommendation that if one is to err, it would be better to err on the side of ETTR. I make that recommendation because my anecdotal experience is that far more photos are presented here under exposed than over exposed.

How they're presented - in most cases - is a reflection of sub-optimal processing - which can be fixed in post-processing. It's not particularly related to gross under-exposure at the time of capture.

Fourth, bringing the context of getting the gamma right is so far beyond Maria's current understanding that it's unhelpful to her.

Yes - it's not something that she should have to worry about just yet - BUT - it's something she's going to HAVE to worry about if she starts following your advice and using ETTR all the time.

Otherwise, your post was dead-on.

Glad we agree

[Inkanyezi]

As has been hinted, the histogram can be a help in finding correct exposure, although it takes a good deal of learning to understand how.

It is a help when post-processing, in order to utilise as much as possible of the dynamic output range of the monitor or when printing.

If he points where the histogram ends are well within the borders, with much baseline between the histogram end and the right or left wall, it clearly indicates that the image is softer than it needs to be, that its contrast can be raised a bit, to get more "punch". The simple way of treating this is adjusting black point and white point, which can be drawn to where the histogram ends. This is the simplest action that can be done also with "Levels", but there are more tricks with curves or levels. The diamonds under the curve diagram in image processing programs represents the "white point" and "black point". Very often we want the histogram to end just there.

Much space with only baseline to the right is a sign that you could have exposed more without over-exposing. Empty space with only baseline to the left at the same time as the histogram climbs the right wall is typical for severe over-exposure. A histogram that climbs the left side, with only baseline to the right, is typical for under-exposure. However, a low key image could have the histogram climbing the left wall when correctly exposed, although the low key image will mostly have ripples instead of pure baseline at the right side. A high key image will typically have a high peak close to the right side, but should, ideally, slope down to the right corner and not climb the right wall, and it can have ripples to the left. So a "peak" far to the right, without hitting the wall, is typical for a high key image.

Everything you do to the tone curves of the image will be visible directly in your image processing program. The histogram does not force you to any action, and there is no particular shape to prefer over any other. However you can see where curve adjustment can have an impact, for better or worse. Lifting the curve where there is a peak, will lower that peak, causing contrast loss above that luminosity and contrast increase in the darker regions. Lowering the curve where there's a "valley" will raise it, causing darker regions to become even darker and increasing contrast in brighter parts.

With time and practice you will learn what happens and what can be done. Changes you do will not always improve the image, but just as swimming is best learned in the water, you cannot learn image processing by words alone.

[chauncey]

I would submit that learning to use that histogram, AKA, getting proper exposure, is somewhat of an art form. Years ago, the maligned Andrew Rodney, pushed/prodded/nagged/guided me in developing my technique of properly ETTR by relying on my LCD RGB histogram, without that feature, I most certainly would be up the proverbial creek.

In my definition, properly ETTR means that you come close to, but do not cross that right boundary with any of the three colors after...neutralizing the in-camera picture style settings as they will affect the jpeg representation on that LCD screen.

Should you not be blessed with having that RGB histogram displayed on your LCD screen, there are other techniques available but, knowing them would accomplish nothing more than cluttering up my limited capacity brain.

[Colin Southern]

As has been hinted, the histogram can be a help in finding correct exposure, although it takes a good deal of learning to understand how.

It is a help when post-processing, in order to utilise as much as possible of the dynamic output range of the monitor or when printing.

If he points where the histogram ends are well within the borders, with much baseline between the histogram end and the right or left wall, it clearly indicates that the image is softer than it needs to be, that its contrast can be raised a bit, to get more "punch". The simple way of treating this is adjusting black point and white point, which can be drawn to where the histogram ends. This is the simplest action that can be done also with "Levels", but there are more tricks with curves or levels. The diamonds under the curve diagram in image processing programs represents the "white point" and "black point". Very often we want the histogram to end just there.

Much space with only baseline to the right is a sign that you could have exposed more without over-exposing. Empty space with only baseline to the left at the same time as the histogram climbs the right wall is typical for severe over-exposure. A histogram that climbs the left side, with only baseline to the right, is typical for under-exposure. However, a low key image could have the histogram climbing the right wall when correctly exposed, although the low key image will mostly have ripples instead of pure baseline at the right side. A high key image will typically have a high peak close to the right side, but should, ideally, slope down to the right corner and not climb the right wall, and it can have ripples to the left. So a "peak" far to the right, without hitting the wall, is typical for a high key image.

Everything you do to the tone curves of the image will be visible directly in your image processing program. The histogram does not force you to any action, and there is no particular shape to prefer over any other. However you can see where curve adjustment can have an impact, for better or worse. Lifting the curve where there is a peak, will lower that peak, causing contrast loss above that luminosity and contrast increase in the darker regions. Lowering the curve where there's a "valley" will raise it, causing darker regions to become even darker and increasing contrast in brighter parts.

With time and practice you will learn what happens and what can be done. Changes you do will not always improve the image, but just as swimming is best learned in the water, you cannot learn image processing by words alone.

I think that - in summary - a histogram is only useful as a comparison tool for the photographer to compare "what they have" with "what they should have". Unfortunately, "what they should have" can be difficult to master, and until it is mastered then trying to apply "one size fits all" advice just makes things worse.

Case-in-point: Yes, often, we will want to stretch the histogram in processing to clip both black and whites and thus utilise a full tonal range - however - there are times when that would be 100% inappropriate (eg a closeup of skin) may have no highlights or shadows at all - in which case a full total range histogram would be the last thing you wanted.

Personally, I use blinkies to indicate areas of potential over-exposure (so that I can evaluate those areas), and I use the histogram to indicate how much of the sensor's dynamic range I'm using (comparing that to the scene before me and other considerations); if it indicates that I still have a couple more stops available to me - but the image looks as expected then "job done"; if it indicates a couple more stops available to me - and I need those 2 stops for a high dynamic range scene - then I'll probably increase the exposure.

[Colin Southern]

In my definition, properly ETTR means that you come close to, but do not cross that right boundary with any of the three colors after...neutralizing the in-camera picture style settings as they will affect the jpeg representation on that LCD screen.

The in-camera generated histogram - even when picture styles are "neutralised" - will still be conservative compared to RAW data. Normally, that's the way I like it because the sensor is capturing more data than necessary anyway, and the inherent safety margin of that technique is a good compromise in my opinion.

I still maintain that ETTR is still mostly a waste of time though; rather than aiming for ETTR all of the time, I'm more interested in matching the dynamic range requirements of the scene to what's available to me in the camera; if it's a shot into the light with a requirement to protect foreground shadow detail then one needs to expose carefully (the result of which being something that resembles ETTR), but if there's nothing challenging in a reflective scene with normal shadow requirements then I wouldn't bother with ETTR personally - it's just more work and risk for only a theoretical (not practical) advantage, in my opinion anyway.

Andrew Rodney has probably been hanging around Jeff Schewe for too long!

[ajohnw]

Sorry, John, but nothing in your post contradicts my post. I wrote that "ETTR will almost always be your better choice, especially at this stage of [Maria] learning how to apply the information displayed in the histogram to the image." In my opinion, you and all the other folks who are very steeped in the nuances of ETTR and the relatively few situations when using it is not appropriate, are overlooking the context of the OP. While your posts are appropriate for some of the others following the thread, the most important posts in the thread in my opinion are those that respond directly to the concerns, experience and context of the OP.

Hello. I'm having some difficult fully understating how to interpret and analyze a luminosity and color histogram. I've read the tutorial but am still a little lost :-) Please help! Thanks in advance!

Please tell me how ETTR relates to the original question Mike as basically it has nothing to do with it. Given what the majority of cameras do it's just a poor example of their use.

John
-

[ajohnw]

The in-camera generated histogram - even when picture styles are "neutralised" - will still be conservative compared to RAW data. Normally, that's the way I like it because the sensor is capturing more data than necessary anyway, and the inherent safety margin of that technique is a good compromise in my opinion.

I still maintain that ETTR is still mostly a waste of time though; rather than aiming for ETTR all of the time, I'm more interested in matching the dynamic range requirements of the scene to what's available to me in the camera; if it's a shot into the light with a requirement to protect foreground shadow detail then one needs to expose carefully (the result of which being something that resembles ETTR), but if there's nothing challenging in a reflective scene with normal shadow requirements then I wouldn't bother with ETTR personally - it's just more work and risk for only a theoretical (not practical) advantage, in my opinion anyway.

Andrew Rodney has probably been hanging around Jeff Schewe for too long!

I generally agree 100% with that Colin but must point out that jpg - raw comparisons these days depend on what jpg mode the camera is in. On some it's a fact that getting more out of raw is rather dubious. Most "standard" modes get 9 stops into the jpg. Some "curious" modes a lot more. They do this via tone curves in just the same way as people develop from raw. Many cameras have several options in this respect. Some leave very little scope for extracting extra highlights from raw in all modes.

It depends on make. Even Canon are reluctant to just chop of highlights in jpg's these days and all are working on processing the dark end in camera on jpg's. Personally though I feel all cameras should have a "Canon" curve in them for use when the scene is suitable.

John
-

[Colin Southern]

I generally agree 100% with that Colin but must point out that jpg - raw comparisons these days depend on what jpg mode the camera is in. On some it's a fact that getting more out of raw is rather dubious. Most "standard" modes get 9 stops into the jpg. Some "curious" modes a lot more. They do this via tone curves in just the same way as people develop from raw. Many cameras have several options in this respect. Some leave very little scope for extracting extra highlights from raw in all modes.

It depends on make. Even Canon are reluctant to just chop of highlights in jpg's these days and all are working on processing the dark end in camera on jpg's. Personally though I feel all cameras should have a "Canon" curve in them for use when the scene is suitable.

John
-

I think what you read wasn't what I was meaning when I wrote "even when picture styles are neutralised" (in reality my working was a bit loose).

Let me put it another way.

- I recommend people shoot RAW.

- If they are shooting RAW then I recommend that they set their picture style to neutral.

- If they're shooting RAW and with a neutral picture style then the JPEG histogram and highlight alerts(s) will be conservative (on Canon DSLR anyway). Generally, evaluative metering - of a black / white (ie 4-stop) reflective object - at base ISO - will give around a 2 1/3 stop safety margin above the highlight (ie you can increase it by around 2 1/3 stops before true clipping occurs IN THAT CIRCUMSTANCE) (other scenes - especially where strong localised backlighting is present - will meter/expose differently). Or "in English", properly metered studio shots (eg light meter) have around a 2 1/3 stop safety margin (having said that though, I'd never even THINK of going full ETTR in the studio - with thinks like skin tones it would be a nightmare trying to get the gamma/midtones exactly where they should be.

By and large - in my opinion - ETTR (and the intrinsic reliance on the histogram to achieve it) - seeks to solve a problem that doesn't exist most of the time. Yes, there are exceptions, but they ARE exceptions.

[Mike Buckley]

Yes - it's not something that she should have to worry about just yet - BUT - it's something she's going to HAVE to worry about if she starts following your advice and using ETTR all the time.

It would help if you would read my post again, as I never advised that she use ETTR all the time. That fact also negates all of your other comments. Even so, this discussion has been good because it's clear to me that you, I, and most important, Maria, understand the important issues.

[Mike Buckley]

By and large - in my opinion - ETTR (and the intrinsic reliance on the histogram to achieve it) - seeks to solve a problem that doesn't exist most of the time.

Your comment makes me think your and my concept of ETTR are different, as I ETTR in almost every photo that I capture and it's not because of a perceived problem. It's simply to achieve the ideal exposure using the histogram as the guide. If you and I have different concepts about ETTR, that would explain at least part of our disagreement about what is best for Maria at this stage in her learning curve.