Exposure Compensation

[pnodrog]

..........

The right exposure is the exposure you get from the lightmeter. And relative from that there is an under or over exposure. And in the extreme clipping can occure.

George

No no no -- The right exposure is the exposure that will best capture the tonal values in a scene. Often a lightmeter will give you the correct exposure indication but for many scenes it will fail to. After all if your camera or lightmeter was always right there would be no point in having the ability to set an exposure compensation value.....

[george013]

No no no -- The right exposure is the exposure that will best capture the tonal values in a scene. Often a lightmeter will give you the correct exposure indication but for many scenes it will fail to. After all if your camera or lightmeter was always right there would be no point in having the ability to set an exposure compensation value.....

That's why I wrote not to start a discussion on the lightmeter. Somewhere there is a "right" exposure and on both sites an under and over exposure.
It all started with 2 different definitions of ETTR from Urban. Than the meaning of over exposure and now the lightmeter. I'm sure you know what I mean.

After all I still don't know the explanation of Urban conserning the different definitions of ETTR with whatever definition of over-exposing.

George

[xpatUSA]

. . .

The right exposure is the exposure you get from the lightmeter. And relative from that there is an under or over exposure. And in the extreme clipping can occure.

George

Well, there we have it, gentlemen, finally explained without obfuscation!

George believes that over-exposure is above a 'correct' exposure which is given by the metering system and vice-versa for under-exposure.

Others are more comfortable with overloaded sensor wells as being the definition.

I myself include the conversion in the process and, from a given raw image, the converted image might or might not be over-exposed. That makes me sympathetic to George's view except for the metering perhaps. Certainly the conversion for Foveon images is extreme in terms of multipliers - such that clipped values in the RGB conversion are quite possible, even when the sensor itself is comfortably exposed. In IR imaging, Foveon converters often force all the greens to zero, even though all three layers respond to IR wavelengths, i.e. there is 'green' data in the raw file and even some 'blue'.

Which leads me to think that the phrase "over-exposed", stated by itself without qualification, means all things to all men - just like those sad words "resolution" and "quality". Therefore should we not simply take the phrase in context and, if confusion occurs, seek clarification?

Case in point, albeit well OT:

I say that my 12MP GH1 has twice the resolution than a 12MP D700. Others say that the resolution is about the same. Let's not discuss whether it is or it ain't - it's just the points of view that differ. Had I qualified "resolution" with "sensor geometric", I would have been right and Others would nave been wrong. Had Others qualified "resolution" as "lines per picture height" they would have been right. Had we all qualified "resolution" properly, we all would have been right! :-)

What say you all?

[Manfred M]

That's why I wrote not to start a discussion on the lightmeter. Somewhere there is a "right" exposure and on both sites an under and over exposure.
It all started with 2 different definitions of ETTR from Urban. Than the meaning of over exposure and now the lightmeter. I'm sure you know what I mean.

After all I still don't know the explanation of Urban conserning the different definitions of ETTR with whatever definition of over-exposing.

George

George - I have been sitting back and biting my tongue a bit here. I think the issue may be down to semanics.

First of all there are two parts to "exposure"; the first is capturing the image in the camera and the second part is reproducing it to be seen, whether that be in print or on a computer screen. The two are intimately linked, BUT exposure is purely a camera function. Here Urban, John and Ted are saying much the same thing; from the camera view, an image can be viewed as being properly exposed so long as there is no clipping of highlights or loss of shadow detail. The obvious exception is of course when the dynamic range one is trying to captures exceeds the camera sensor's ability to do so. A modern camera, at base ISO can capture over 12 eV or more of dynamic range.

The camera lightmeter is a bit of a red herring, as all it does is ensures that your image capture is going to give you that fabled 18% gray reading (and even here, depending on your source, that 18% is being debated ad nauseum). We already know that this is not going to work for all scenes; a snow scene, a night scene, etc. are not 18% gray, so we will compensate by either using the exposure compensation setting on our camera or shooting manually.

When I shoot in a studio like setting with studio lights, I don't rely on my in-camera exposure meter, but rather use an incident light meter that measures the amount of light hitting the subject, so the 18% issue does not apply. Even here, I may chose to overrule my meter reading to give me a result that "works better", if there is something a bit funkly about my subject, background and lighting.

That brings us to the real issue; the output. All of our output devices, from paper print to computer screen have a far lower dynamic range than what our camera is capable of capturing. These devices end up compressing the captured image so that it can be displayed. Again, the exact numbers are up for debate, but from what I remember, a print has around 4 eV, a LCD type screen around 5.5 - 6 eV and the old CRT displays could handle up around 7 - 8 eV.

If you are judging "correct exposure" based on how the printer or display software decide how to output your image, based on the output compression algorithm used, then yes, the image may look over or under exposed. If on the other hand, you understand this and use the additional "headroom" available in your unclipped / unblocked data (RAW) file, to set the compression to where it should be, then no, your image was not over or under exposed, it was incorrectly processed by the computer / camera software.

ETTR (and other techniques) use this difference between what your camera has captured and what you determine the correct output to be. In cases where there shot you are trying to take does is less than the dynamic range that your camera sensor is capable of capturing, there is a good technical reason to bias the exposure towards the highlight end (ETTR) is order to reduce the amount of digital noise which is more apparent in the darker parts of the image.

So what you see in your output isn't necessarily over or under exposure. It's the result of how the image was processed that can be overly light or overly dark. As long as you have enough headroom in the data, manually overriding the software is the right thing to do.

[Inkanyezi]

That's why I wrote not to start a discussion on the lightmeter. Somewhere there is a "right" exposure and on both sites an under and over exposure.
It all started with 2 different definitions of ETTR from Urban. Than the meaning of over exposure and now the lightmeter. I'm sure you know what I mean.

After all I still don't know the explanation of Urban conserning the different definitions of ETTR with whatever definition of over-exposing.

George

Then, if you don't understand, I think it would be wise not to be so hard-headed about your simplistic definition, but instead listen and try to understand, als je begrijpt wat ik bedoel.

You are simply wrong. Totally wrong. With your simplistic view, we could not have evaluated that a negative was under-exposed in the past, as we could not assess what the light meter would have shown. Nevertheless, in those days we regarded a negative that was too thin to print an image, as under-exposed, and we regarded the blocked up negative as over-exposed, because the highlights could not be rendered well.

If you read the Ansel Adams book The Negative, you might begin to grasp that "right exposure" is not such a simple concept. Correct exposure will depend on your workflow, whether you use silver salt technology or semiconductor sensors. Ansel Adams used the spot meter to find his best exposure and workflow, and his zone system could be applied also with the digital technique. However it has a few drawbacks, which might not be so apparent when you use film, but with digital, we can improve our method, by another workflow.

In the zone system, all brightness values are assigned zones, and development can expand or contract the scale of possible values. Digital is different, as we are actually not very interested in the lowest values that our digital gear may give us. They are full of noise. Therefore, we may apply a different idea about the best workflow. If we apply the zone system similarly, we could adjust our dynamic range, by setting an ISO value that will supply just as wide DR as needed. However, we can also use a more narrow part of the wider DR of a lower ISO setting, and the advantage is that we won't get noise in the shadows.

It all boils down to the workflow that is chosen.

We may simplistically use the jpeg conversion of the camera and think that its meter gives us a "correct" exposure. In many cases this will work fine, and for many people thinking gets no further. When using a cellphone to take the images, one is usually satisfied by the result that the all-automatic exposure and conversion does. Under such conditions, one may well think that "correct" exposure is what the light meter would give you.

When exposing to the right and not raising ISO, as we can do when our subject is still and we use a tripod, or when there is ample light, so we won't have to worry, we will not use the far left of the RAW histogram, but expand from the right, so that the histogram will cover more, so that we get our blacks. We also have the possibility to shrink the dynamic range to fit the histogram, thus gaining possibility to use a faster shutter speed or a smaller aperture or both. This is then done at the expense of more noise in the shadows.

Exposing to the right is not always necessary, but its advantages when ISO is higher are evident.

So "correct exposure", when applying ETTR is to pile up the bulk of the histogram toward the right wall - i.e. the RAW histogram, without clipping any important data. This will certainly not be what the light meter shows. In most cases it will be more, and in a few cases less than the camera's meter suggests.

And the stupid light meter without even a brain is not going to tell me what "correct exposure" is. I am the one that must make the judgement, because I am the one that takes the picture. The machine isn't.

You can easily see, that the light meter will not give you the correct exposure, by taking an image of a snow field in the sun, with a few people in it and some vegetation, say pines. When the scene mostly has snow in it, the light meter will give you a way too high reading, which must be corrected, for the image to come out correctly exposed. It is a situation easier than red flowers, as your snow is rather neutral in tone and will be assessed correctly by the light meter. But the light meter can only evaluate its brightness, not how it shall appear in the image. You have to make a correction to the reading, for the brightness you want the snow to have in the image. If you don't do that, the pine trees will be black and the people in the snow also will be black. The snow itself will be dark grey. The image will be under-exposed, if you use the reading without correction.

If you use another system for measuring light, the meter may give you a totally different suggestion of how to expose. In the movie business, the preferred way of measuring light was to assess the incident light, the light falling upon your subject. Mostly this method will give better results provided you use standard development, as it assumes a standard development and that all reflective objects shall be rendered according to their reflectance. The method works well for digital too, but it was not the method of choice for Ansel Adams, as he would alter his dynamic range in PP.

So your error is the concept that the meter will give you a correct reading for exposure, and that any deviation from that would be either to over-expose or under-expose. That idea is wrong. The meter gives a value, but it is up to the photographer to evaluate whether to use it as is, or to alter it according to image content and workflow. ETTR calls for a workflow adapted to the technique, and it does not suggest over-exposure. ETTR has a clear definition of "correct" exposure, which is the exposure that supplies the most data without clipping any of the channels. It is far more stringent than the one you propose: "The right exposure is the exposure you get from the lightmeter."

Some day perhaps a light meter can be invented that does just that, gives you the right exposure. Till then, we have to tweak the suggestions our light meters give us as we need to. Misschien hoeft men een groter denkraam om het te begrijpen?

[xpatUSA]

After all I still don't know the explanation of Urban conserning the different definitions of ETTR with whatever definition of over-exposing.
George

I think ETTR is yet another photographic term which, by itself, means different things and should really be qualified. Some people do - and some people don't - hence confusion reigns.

ETTR = expose to the right. To the right of what, one asks? "the histogram, of course, fool".

Um, what histogram, one asks? "Oh . . . er . . well . . the one for the image, duh".

Which image? . . etc., you get my drift . . .

And then there is the scene-dependence of the 'correct' exposure as has been already explained. So any kind of metering, spot, center-weighted, multi-point evaluated or even a hand-held thingy, has to be interpreted - not just accepted as-is.

Pardon my rambling on . . sorry.

[Manfred M]

Folks - let's tone down the sniping a bit.

There is a lot of good information on this thread, so my issue is not with what is being said, but how it is said.

Please, when responding, please re-read your posts and clean them up if the response might not be in the best tone and try to not make it personal. Lets try to keep the questions and responses at a professional level.

We are all here to learn and share knowledge. I do understand that there are members whose first language may not be English, so writing at an appropriate level should be considered as well. Basic understanding of the technical aspects will vary too and will be read by others.


Thanks

Manfred

[Polar01]

George now matter what a lightmeter is involved either the one in your camera or a handheld one. To me I have a simple definition of over and under exposed, over is in the highlights areas where there is no data collected by the sensor, under areas in the darks where no data is collected. I do not worry about individual channels, or the histogram that much as what is showing on the camera is taken from the jpeg that is created by the camera to view on the view screen. As a jpeg I could care less as it is the raw capture that it more important to me that I take into post production. I find in post that channel that was clipped is not clipped, the warning light that the camera displayed in the highlights for clipping not there in post.
Do I look at the histogram and blinkies, yes they are only a guide they are telling me that I am close to what it guesses is the correct exposure for the scene. If I take a shot, and the histogram and blinkies tell me that I have clipped highlights, knowing my camera I have at least 2/3's to a full stop or more of exposure before any problems in post.
I look at exposure this way, in the days of film 1st exposure was the time the film was exposed to light by the camera, 2nd was (now some would not agree here) the exposure of the film again this time to chemicals to develop it to what was desired. This could have been just a standard development, or the film was pushed or even pulled. Next would be taking that developed film and using light and now exposing photographic paper (3rd exposure) and the 4th exposure that of the paper to chemicals so that a image can be seen. Now the last two, maybe done a number of times until the desired image is gotten.
So exposure compensation in camera is just us adjusting the camera setting to what we want, not what the camera thinks is correct, and in post exposure compensation is simply moving the exposure slider(now some think that is should be called the "make it darker/make it brighter slider", but exposure works for most of us) to get the look we want.
So is the right exposure in camera correct or is the exposure gotten in post correct? Answer in camera yes/no, post yes.

Cheers: Allan

[Saorsa]

Some day perhaps a light meter can be invented that does just that, gives you the right exposure. Till then, we have to tweak the suggestions our light meters give us as we need to. Misschien hoeft men een groter denkraam om het te begrijpen?

There is NO right exposure.

Is this the right exposure



or this?



Both were taken in bright Florida sunshine and exposed as I wanted them to be.

[xpatUSA]

So "correct exposure", when applying ETTR is to pile up the bulk of the histogram toward the right wall - i.e. the RAW histogram, without clipping any important data.

Urban, we were typing at the same time. May I add that your statement is less applicable to Sigma cameras, although it does apply to an extent. The conversion matrix from camera space to XYZ space is the reason why.

Here's one:



Of particular note is the 'blue' (top sensor layer) to Z multiplier of 4X.

Going from camera space to RGB space is, if anything, worse:



This matrix is from a simple X3F-to-TIFF reader utility. A little study shows that a shot with Swedish (i.e. Northern) blue sky in it (comfortably exposed but with the said sky ETTR'd in RAW) is going to have it's converted RGB blue channel blown to smithereens and, in fact, George's metering method might have fared better, IMHO.

[ashcroft]

Have photographers lost the art of looking at a scene and judging exposure?

No. Or at least I haven't. Any given scene/subject can be exposure set and shot to get a realistic look to the finished image. Most people looking at a shot like that would agree that the exposure is 'good'. I'm OK with looking at standard scenes like that and deciding what is required, as are most photographers (but not all).

Alternatively, there are situations where you want to either be more creative with light, or (and this is quite often) there is not enough light, or it's in the wrong place in the shot. That needs more practiced judgment and skill. Someone looking at a creative shot like that might think it's not a 'good' exposure, but the photographer might be wanting to create something different. I have a good example of this. The two shots attached are of the same object. One is a standard product shot and the exposure is 'good'. The other is deliberately more creative using different lighting and exposure. But I'd still say it's a 'good' exposure, but it's different. The first is realistic with a mix of tones. The second is much stronger contrast mostly blacks and whites.

If you are referring more to general shots that are badly exposed, then yes, I think quite a few photographers don't seem to even know when they have got the exposure wrong - even after post-processing and posting on-line. I find that quite puzzling. I see some shots on-line that are clearly off-exposure (I'm talking standard scenes here like a non-creative city landscape scene), and I can tell what the histogram is going to look like.

[ashcroft]

That's a good example of what I was saying above. Looking at the histogram without seeing the shot it might look a bad exposure. But it isn't. If you start making it brighter on the right it soon starts to fall apart. BTW - very nice shot.

[ajohnw]

Some basics.
With exposure you control the amount of light that hits the sensor/film. It's done by a combination of shutterspeed and aperture diameter. Very simple and basic.
The lightmeter inicates what settings should be used for the existing light and sensibility of the sensor/film.
Overexposure is using a higher exposure as the lightmeter suggest.

George

LOL. A whole page more since a posting slip up thanks to a call regarding a car service that caused me to dash out - without actually posting it. This was around 9-30am. I thought it might finish the thread. Thanks to autosave here it is.

Exposure straight off the meter is aimed at jpg's not the sensor. There is also a fair amount of software involved in the transfer of one to another that may do all sorts of things including altering the calibration of the meter.

There seems to be a semantics issue here along with something else that I wont mention. Exposure boils down to several issues. Jpg's, jpg's clipping and single to all channels clipping on the sensor. All are different and have different implications as far as images are concerned. Also workflow. I don't see any real problems with Urban using the word exposure to cover the lot. The main aspect of an EXPOSURE is that it is as correct as it can be subject to the usual limitations bearing in mind the result that is wanted.

John

- Main agents looking for work.
-

[george013]

That is the virtue of UNI-WB and ETTR. If you can use two stops more exposure, you will have lots of more data in the darker parts of the image.

The idea of ETTR and "other concepts like that" is NOT to over-expose. The idea is to expose correctly for the workflow that is chosen. ETTR in no way promotes over-exposure. That is a total misconception of the idea. ETTR means that you place the white point when you expose the image.

The only thing I want to know is how you can combine those two definitions.
In the first one here you state that you have to add 2 stops more exposure, I will use your words, to gain extra info in the darker aerea. I can live with that. It's one of the two aspects of ETTR.

In the second definition you clame that ETTR is placing the white point when you expose the image. That's someting as placing the whitepoint in the correct zone.

I just can't combine them.

By the way, in the tutorial here it's also called overexposed without clipping.

George

[Stagecoach]

George,

You are clearly reading things incorrectly;

The only thing I want to know is how you can combine those two definitions.
In the first one here you state that you have to add 2 stops more exposure,

Here you say that the statement you are referring to states ""you have to add 2 stops more exposure""

But, what the statement actually said is;

""If you can use two stops more exposure,""

You have changed an IF to a HAVE

Why?

[george013]

What a silence. Must be dinertime.
George

[pnodrog]

.......

By the way, in the tutorial here it's also called overexposed without clipping.

George

Good point. It should really be called "increased exposure without clipping" It just highlights the unfortunate fact that the term overexposed can be used in different contexts with different meanings. All that without even referring to sunbathers on the beach....

[george013]

George,

You are clearly reading things incorrectly;



Here you say that the statement you are referring to states ""you have to add 2 stops more exposure""

But, what the statement actually said is;

""If you can use two stops more exposure,""

You have changed an IF to a HAVE

Why?

Change the have in can. It doesn't change the question. What he is saying, if you have space for 2 stops overexposure you should do that to gain more info in the darker part of the image.
The fisrt definition is a dynamic one, you have to change, and correct, the exposure if possible, the second is static, you have to place the white point directly at the right place.
I'm waiting for an explanation.

George

[xpatUSA]

A lot of talk in this thread, much of which makes sense and some that does not, about "the histogram" and "the image" and "over-exposure". Without disagreeing with anyone, I would like to present some images that are less pretty than hitherto but are rather more indicative of the need for interpretation of the scene. Today, I shot a scene of work-in-progress with no particular regard for any particular method. (For me, my shots with the on-camera flash are basically a crap-shoot - often literally.)

I'll start with a conventional linear histogram from the raw file:



Since "clipped" in my Sigma SD14 files is around 9000, the image is obviously under-exposed, is it not? Or is it? Well, if we look at a histogram where the Y axis is logarithmic, a different story appears:



Now we see that clipping in the raw data has occurred in apparently scary amounts, leading us to now believe that the image is badly over-exposed!

As has already been stated, conventionally-scaled linear histograms can be misleading but so can non-linearly scaled histograms unless we know how to interpret what we see.

And so we come to our actual raw image:



Now it is clear that the capture is mostly "under" exposed with highlights only in the warehouse-style lamps. And sure enough, opened up in Sigma's proprietary converter, we get blown levels:



So, if I had exposed more "to the right" the room would have been better exposed.

Fortunately for me, Sigma's Fill Light is second to none because it both brings up the shadows and recovers highlights all with one slider . . outstanding:



If anything, this post illustrates the need for scene interpretation quite well.

RawDigger - don't leave home without it.

Sigma Photo Pro - still the best converter for X3F files, IMNSHO

[Inkanyezi]

The only thing I want to know is how you can combine those two definitions.

Neither of them is a definition. Re-parse the text in the quote, and you will find an "if" in it. It says "if", not that you should necessarily do so. But if you can expose a couple of stops more without blowing highlights, you will have more data for your PP rendition of the image.

In the first one here you state that you have to add 2 stops more exposure,

That is exactly what I do not state. Re-parse the text, and you see that I don't say that. What I say is this:

If you can use two stops more exposure, you will have lots of more data in the darker parts of the image,

I hope you can see that the sentence starts with "if".

I will use your words, to gain extra info in the darker aerea.

That as well is something I did not state, even if it is true. Your interpretation was luckier here.

I can live with that. It's one of the two aspects of ETTR.

It is in the very heart of the idea, that you should have ample data to play with in PP.

In the second definition

Mind that it is not a definition, it is a sentence that you isolate from its context.

you clame that ETTR is placing the white point when you expose the image. That's someting as placing the whitepoint in the correct zone.

True, you place the white point in its correct zone - or you can place any other point of your choice, into its relevant zone. When your image will not have any white at all, you surely would place the most exposed pixels in another zone.

It should also be noted, that the zone system is defined for B&W imaging, and we mostly use colour in our digital imagery. The zones are not quite relevant as originally designed. A red that is close to clipping is not as bright as a white that is close to clipping, but still, it must be held within the limits of your representation of the image, the file, in order to be rendered well in the presentation of the same.

I just can't combine them.

That is a comprehension issue, but I have tried to get it through logically.

The workflow is such, that the brightest parts of the channel that is closest to clipping are rendered as bright as the presentation medium will allow. In the case of red flowers, it is not a white highlight, but a red tone, close to clipping. It is placed in its correct zone at the moment of exposure. As it is not white, but a colour, it will not be the white point. White point is valid for white objects, as clouds that are illuminated by the sun. The red rose is more likely to fall somewhere closer to zone VII. but if it had more exposure, it would clip. So "white point" in the phrase was a simplification. But I did not want to make it more complicated. Still, the tonality is decided at the moment of capture, and you may express it in tones of the zone system.

If you choose another workflow, as for example the one of creating a jpeg in the camera, the image could look as if it were over-exposed, but it isn't - it is just processed in the wrong way. In older days, we had the concept of "over-developed" also. When the zone system is applied, what we wanted to do with a low contrast scene was expose less and develop more. When this process was chosen in order to place the scene tonality in the desired zones, we would not consider the negative under-exposed or over-developed, but correctly exposed and developed for the chosen method, the Zone System.

If that negative would have been processed as suggested by the tables in the developer manual, it would appear under-exposed, because it would not be developed as intended.

The light meter is just an instrument. A resourceful photographer can use it to get correct exposure for the workflow that is intended. If that workflow comprises a standard jpeg processing with standard curves. It is unwise to use ETTR techniques and UNI-WB. If however the intended workflow adheres to the ETTR principle, UNI-WB can be used to see that the brightest channel merely touches clipping without going there. Processing is then done to get the best possible image out of the RAW file.