Limits of Exposure Composition

[Abitconfused]

This image, which has not been altered, has no black point and no white point. It occurs to me that exposure composition alone (+ or -) could not have remedied this exposure to provide a more compelling contrast. At this point, should the photographer meter from the foreground, apply exposure lock, and recompose the image in an effort to obtain a suitable exposure? Or, perhaps, use spot metering on the buoy?

[Shadowman]

I use EC to alter settings chosen by my camera when shooting in A or S mode. Usually I'm trying to bring back some detail in a blown out background, tightening up small highlights. You should be able to find this capture's B/W point in ACR, just hold down the ALT key and move the black slider until you see becoming visible.

[davidedric]

Hi,

The scene basically lacks contrast, so changing the exposure (which is what ec is) won't make a lot of difference. Think of it as moving the histogram one way or the other, but basically not changing it's shape. From the image and the histogram, it appears to be a little over exposed, but that's obviously easy to fix. Then you need to work on adding contrast, including black and white points. I'm not familiar with Photoshop, but there are several different ways of increasing contrast in Lightroom, so there must be in ACR (are you shooting RAW?), so it's a case of using whichever gives the effect that you are looking fo.

Apologies if this is all a bit obvious!

Dave

[dje]

Yes I think this is a case of a scene with a small dynamic range, rather than an exposure issue. It just needs a Levels adjustment of the black and white point to stretch out the tone range, together with a contrast increase.

Dave

[Dave Humphries]

Hi Ed,

Just to add a third Dave in to this discussion

As the two Dave's have said, changing exposure (or your metering choice) will not affect the issue in any way.

This is why any/every image needs PP.

If you're not shooting RAW (i.e. you are shooting jpg only) and want to get it right SOOC (straight out of camera), then when shooting this kind of scene and seeing that on the camera's histogram, you need to go in to the camera's menu and increase the contrast used - but that's a lot of hassle on site and it is soooo easy to forget to restore it when the sun comes out and scene contrast increases - then you'll end up with too much contrast and permanently lose detail in your highlights and/or shadows

Thus, it is far safer to shoot RAW and deal with it in Post Processing (if you're not already).

To fix:
In that screen shot above (Levels Dialog), I'd grab the little LHS triangle under graph and move it left to bring the black point up to where the pixels are in the histogram, then grab the right hand one and move that left to meet the pixels at the 'white' end. That'd would be a start anyway - although there are other techniques which would result in a better final result for the overall image.

Cheers, Dave

[Mike Buckley]

Nice composition, Ed! To clarify just a couple details about terminology...

The term is exposure compensation, not exposure composition. When using that capability, you are compensating for the meter's reading of the luminosity values in the scene.

You mentioned that the image has no black and white points. Every image has black and white points. It's a matter of changing them when desired, which is the case everyone is explaining about this one.

[Thlayle]

Ed,

Your image reminds me of a common problem I find taking pictures in Michigan much of the year. At some point in the fall, the color palette reduces to a muted range and the skies turn to an amorphous grey. This lasts for about five months and is abated only by either the seasonal snow cover that gives some different challenges or abnormally warm seasons in which the landscape has more variation to the color range with more things being green in winter (and more sun).

I think the comments offered already about the lack of contrast & limited dynamic range speak to the real issues & challenges for this kind of setting. I am sure some others may have some ideas about getting this right at the point of capture. For me, other than a road trip for different shooting settings, I depend a lot on post processing to get more from some of these situations.

[DanK]

The Daves are spot on, but it might help to recast this in terms of what the histogram shows. If this is stuff you already know, excuse me.

As you probably know, the histogram shows the distribution of luminance, or brightness. the left-hand edge is pure black. The right-hand end is pure white. The height at any point is more or less the proportion of pixels in the image that are at that level of brightness.

Any exposure change from any cause, whether it is the use of EC, a change in metering, or a change in manual settings, affects the location of the histogram--its position left to right. In other words, it just affects the overall brightness of the image. The only time it will substantially change the shape of the distribution is when you clip--over- or underexposing so that the one end of the histogram bumps up against an end.

The impression of constrast is determined primarily by two different things. One is the total spread of luminance, that is, the tonal range. This is why you have been given the suggestion of moving the white and black points, which will expand the tonal range. The second factor is where in the distribution the transition from dark to light occurs. We perceive more contrast when the transition is concentrated in the midtones. That's why another standard way of increasing contrast is to impose an S-shaped curve using the curves tool. This compresses the lights and darks and spreads out the midtones.

So, the bottom line is that you need to keep in mind two separate things: the level of luminance, which is controlled by exposure, and the distribution of luminance, which is controlled by various tools, ideally in software after the capture.

[Mike Buckley]

Great explanation by Dan! Not so much detail that it becomes overwhelming yet just enough detail to pique the interest of anyone wanting to improve any image displaying the characteristics being discussed.

[Manfred M]

Let me just add to the discussion as to when exposure compensation is required when shooting.

To properly assess exposure, we should be measuring the amount of light that falls on our subject / scene (incident light). This is not possible when we build a light meter into our cameras as the measurement is taken of the light that is reflected from the scene (i.e. reflective light meter). In order to make things work, the camera's light meter has been set up to measure what happens in a scene that reflects an average amount of light. This generally works quite well in many scenes, but when the scene is not average, we can overrule our light meters to compensate.

I use exposure compensation a lot during Canadian winters when there is snow on the ground, as the light meter is fooled and tends to underexpose, so I tend to dial in between +1 and +2 stops of exposure compensation. I get the same types of issues when I shoot at night, again, the exposure meter is set to average things out and these scenes will often be overexposed and turn out too light, so negative exposure compensation needs to be dialed in. I find that a -1 to -2 stops work here as well.

In both these cases, I study the histogram from a test shot and dial in enough compensation that gives me a good histogram to work with. Night shots are tricky as lights can (and sometimes should) blow out in the image, so I sometimes misread and don't dial in enough compensation.

When I look at your histogram, the exposure looks good. Some of the waves are close to pure white and some of the wet rocks are getting close to black. I would probably do what others have suggested and push the image a bit further in post, but not much. As for exposure compensation, I suspect that even adding +1/3 stop might be too much.

Unfortunately, those gray overcast conditions where the cloud cover looks like a textureless mass tend not to be great for the type of shot you have here.

[xpatUSA]

Nice composition, Ed! To clarify just a couple details about terminology...

The term is exposure compensation, not exposure composition. When using that capability, you are compensating for the meter's reading of the luminosity values in the scene.

You mentioned that the image has no black and white points. Every image has black and white points. It's a matter of changing them when desired, which is the case everyone is explaining about this one.

Pardon my pedantry Mike, but I think that the word 'level' is preferable to 'point' because the phrase 'white point' is also often used to describe the white reference value used in white balancing. And the terms 'black point' and 'white point' are both used in ICC profile viewers and do not refer to tone levels.

[Manfred M]

Pardon my pedantry Mike, but I think that the word 'level' is preferable to 'point' because the phrase 'white point' is also often used to describe the white reference value used in white balancing. And the terms 'black point' and 'white point' are both used in ICC profile viewers and do not refer to tone levels.

Also my thought Ted, but for a slightly different reason. When editing I will often set a black point and white point during my edits, where everything to the left of where I set my black point will be converted to a black value (0). The opposite is true for the white point, where everything to the right of where I put the white point will be give the value of pure white.

The resulting edit then redistributes all the tonal values between these two extreme and the resulting image covers the full tonal range from pure black to pure white.

[Mike Buckley]

Pardon my pedantry Mike, but I think that the word 'level' is preferable to 'point'

You might be correct, Ted. It's truly impossible to know because the two terms are so often used interchangeably and because so many of the terms used in photography are not standardized.

[Mike Buckley]

I think that the word 'level' is preferable to 'point'

Also my thought Ted...When editing I will often set a black point and white point during my edits

I'm confused, Manfred. If the term, level, is preferable, why did you then use the terms, black point and white point in the same context I used them?

[Saorsa]

I snagged the screen capture and cropped it to just the image.

Then I opened it in Capture NX2 and put a white point at the broadest part of the wave above the red thing on the edge of the sand and a black point on the rock. This is what I got from that simple fix.



Those individual points may or may not be exactly or 255,255,255 or 0,0,0 but they are the whitest and blackest points I could find easily. I sometimes use the pupil of an eye or the black inside the wheelwell of a car.

What these points do is establish a limit for new black and white and the software then interpolates the intermediates relative to the 'new' black and white. In fact, if you look at a historgram after doing this you no longer see a smooth curve. There will be gaps where no value is used.

[Manfred M]

I'm confused, Manfred. If the term, level, is preferable, why did you then use the terms, black point and white point in the same context I used them?

Mike as Ted pointed out in his post, we are getting a bit pedantic here, but none the less, the statements are correct. Engineers. physicists, mathematicians tend to look at things from a more technical view than most photographers.

I'm going to stick to the RGB colour model and 8-bit for simplicity's sake, but the argument extends to other bit depths and colour models precisely the same way.

A point is a discrete entity. In RGB black has been assigned the integer values of (0, 0, 0). A point with the value of (1, 0, 0) might look pretty black, but in reality, it is very, very dark red. (1, 1, 1) would also likely be mistaken for a pure black, but again, no, it's a very, very dark gray. So when someone refers to a black point, to me that means he or she is referring to only one possible set of R, G and B values.

From a practical point when viewing an image pure black and one of these other values that are "almost" or "near" black are indistinguishable from black. This is where levels come in, as these do not have a precise meaning of value. When we are looking at setting a black point or white point we are making a completely arbitrary judgement call as that we are going to let all values on the left side of the black point on the luminence histogram to be set to black and all values on the luminence histogram to the right of where we have set the white point to be set to white (255, 255, 255), for 8-bit white RGB.

Clear as mud?

[Abitconfused]

Perhaps oversaturated and oversharpened but different.

[Mike Buckley]

That makes sense, Manfred.

However, let's get back to the OP's thinking that the original image has no black or white point and my thinking that all images have a black and white point. Let's also get back to my assertion that these terms have not been standardized. (I look forward for someone proving me wrong about that because having a standard to refer to would make life so simple.) Now also consider that, as an example, Lightroom has a Whites slider and a Blacks slider. Note that that software is probably used more than any other image-editing software. Those sliders clearly indicate a wide range of points on the Whites slider (many white points) and a wide range of points on the Blacks slider (many black points). So, until someone can show me at least one standards body and hopefully more standards bodies that has defined that there is just one black point and just one white point, I'm not willing to accept that definition as an effective definition everyone should use.

I would add that the terms are also used inconsistently by the same person, myself included. As an example, by your definition, there is no such thing as changing the black or white point because there is only one black point and only one white point.

The only thing that matters is the essence of what we're trying to communicate and that we understand each other relatively effectively.

[Saorsa]

That makes sense, Manfred.

However, let's get back to the OP's thinking that the original image has no black or white point and my thinking that all images have a black and white point.

It's possible that they might have been lost in processing but when I extracted the image from the screen cap I could not find a pure white or black point. I increased magnification to the point that pixels were distinct squares on my screen and examine the 'white' of the surf and the black of the large rock and vehicles on the shoreline. I chose the whitest of the surf and the blackest of the rock and established a white point for the software. The result is shown above

[xpatUSA]

That makes sense, Manfred.

However, let's get back to the OP's thinking that the original image has no black or white point and my thinking that all images have a black and white point. Let's also get back to my assertion that these terms have not been standardized. (I look forward for someone proving me wrong about that because having a standard to refer to would make life so simple.) Now also consider that, as an example, Lightroom has a Whites slider and a Blacks slider. Note that that software is probably used more than any other image-editing software. Those sliders clearly indicate a wide range of points on the Whites slider (many white points) and a wide range of points on the Blacks slider (many black points). So, until someone can show me at least one standards body and hopefully more standards bodies that has defined that there is just one black point and just one white point, I'm not willing to accept that definition as an effective definition everyone should use.

I would add that the terms are also used inconsistently by the same person, myself included. As an example, by your definition, there is no such thing as changing the black or white point because there is only one black point and only one white point.

The only thing that matters is the essence of what we're trying to communicate and that we understand each other relatively effectively.

Taking Adobe as a de facto standard, I found a page that describes their Levels function:

https://helpx.adobe.com/photoshop-el...elsadjustments

(or, should I call it the Points function? Just kidding, folks)

There, the word "level" appears 23 times.

The word "point" appears once only, in a Macintosh note:

You can click Auto to move the Highlight and Shadow sliders automatically to the brightest and darkest points in each channel. This is the same as using the Auto Levels command and may cause a color shift in your image.

Was going to make another point (!) earlier. It has been stated that a "point" has a unique set of RGB values. If "point" in that context was referring to an arbitrary level in a histogram, the statement is incorrect - at least for luminosity histograms. There are many combinations of RGB that qualify for any particular level in a histogram. We should all know that, so maybe I've misunderstood the statement?