Compromising Exposure in Backlit Situations on Digital

retinax

Well-known
Local time
7:26 PM
Joined
Aug 7, 2015
Messages
1,622
Hi all,

Unless we use fill flash or come back when the light is better, we can either give minimal exposure to avoid any blown highlights. That would probably be ideal, but sometimes the shadows end up so dark that pushing them up in post gives a lot of noise and worse. Even if not, pushing the shadows up gives me very weird colors often, it works better for black and white images.
So what do you do in these situations? Expose for the darker areas and blow highlights completely, or somewhere in between, so that the shadows can be recovered and blow as few areas as possible, or strictly avoid blown highlights?

I'm interested in how other people deal with these situations. I always read about how bad blown highlights are, but I wonder if it's really that dramatic if it's only a dull sky or parts of clouds that get blown. I'd especially like to hear from people who regularly do high quality prints, because I don't (yet), but I wonder if my files are good for it.
I'm talking about daylight photography here, when I'm at or near base ISO, low light photography is a different question.
Best,
J.
 
It depends on the camera for me. The Sigma Merrills even when you blow the highlights seem to transition into the over exposed areas much more gently and they never bothered me too much whereas my M240 is the opposite IMO. An over exposed area just seems to look like a gaping hole with the Leica sensor.
 
Even though modern digital cameras have amazing dynamic range, and ACR can pull detail out of highlights that should have been blown out, you still have to make a decision about where you want to lose detail in some situations. It was the same with film, though. Shoot for the highlights, or shoot for the shadows. Film or transparency film.

Kodachrome shooters (David Harvey is a good example here) usually chose to save the highlights, and let the shadows go black. Until we have perfect cameras with unlimited dynamic range, you take your choice and pay your price.

Most of my photography involves people (PJ), so if I can't use fill flash, I'll blow out the highlights in a bright background and expose the people correctly.
 
Thanks for the replies so far! What happens when you print, Keith? Does the "gaping hole" get better or worse than on screen? And with the Merril, is it fine on screen as well as on prints?
I'm asking because in my own images, the transition to blown areas looks just fine to me on screen, as long as I don't try pulling it down too far in an attempt to recover highlights. If it gets on the steeper part of the (virtual) curve, the transition to blown highlights gets ugly of course. (So Keith, it might help to actually push the transition up the curve a bit, i.e. lowering the contrast in the brightest bits of the image on your M240's files?)
I'm wondering, however, if people find it to be more problematic on prints.

P.S. For my Pentax lenses, I have recently bought an orange filter, that should mitigate the problem with overexposed skies at least when shooting black and white. Unfortunately the GR doesn't take filters...
 
For a single exposure one's options are limited.

If the shadows are actually that dark (- 3 or 4 stops) sometimes it's best to just let shadows be shadows.

Often the all brightest highlights are unimportant to the scene. Letting those highlights go inevitably increases the shadow regions' signal-to-noise ratio.

When the all the highlight regions are important to the scene one must make full use of the analog dynamic range of the camera's data stream. The only way to do this is expose at the camera's native ISO (100 or 200). All other ISO settings reduce the maximum analog dynamic range recorded when the shutter is open.

"I'm interested in how other people deal with these situations.

o I use a camera with a high signal-to-noise ratio. The maximum analog dynamic range is directly proportional to the SNR. It is common for people to use a term such as "high ISO noise performance" instead of SNR. When the scene's dynamic range is difficult "high ISO noise performance" is not a luxury or complying with a gratuitous fad.


o I record raw files and use the camera's native ISO or the lowest practical ISO above the native ISO. My camera is psuedo-ISO invariant which means the shadow noise is almost independent of ISO. Sometimes this means the scene is underexposed. However pushing the brightness selectively during raw rendering has little or no disadvantage with psuedo-ISO invariant cameras. Of course in some cases on-camera LCD review is compromised or impossible.

o When possible I automatically bracket exposures in 1/3 stop steps. This lets me choose a single raw file with the most appropriate highlight overexposure (or no overexposure). In dynamic circumstances and, or at longer shutter speeds bracketing has an obvious disadvantage. Don't confuse this technique with exposure blending. I only use a single exposure with the best possible highlight region exposure. Exposure blending 1/3 stop exposures is almost always a waste of time with my camera.

o During post production rendering I occasionally apply noise filtering (commonly called noise reduction)selectively to shadow regions. This can be useful for large prints.

Here are two links to objective, data-driven comparisons of the analog dynamic range for a large number of cameras. The data is generated from statistical analysis of un-rendered raw files.

These data show different cameras benefit from different ISO and exposure strategies (just as Keith surmised!).

Dynamic Range - Just select the cameras you want to compare from the panel on the far right side of the page.


Shadow Noise Improvement - This data assumes highlight region information will, or can be compromised in order to optimize shadow regions' SNR.

For some cameras the shadow-noise level is dependent on ISO even though the total analog dynamic range decreases as ISO increases. The shadow-noise improvement data only addresses the effects of analog amplification on shadow-region SNR. When shadow regions are important, this data is a useful guide to optimum ISO selection.
 
Hi,

The only thing I like about those little pop-up flash guns on digital cameras is that they are great for a fill-in light if cut back by half a stop. And that's how I cope with the problem...

Regards, David
 
Hi all,

Unless we use fill flash or come back when the light is better, we can either give minimal exposure to avoid any blown highlights. That would probably be ideal, but sometimes the shadows end up so dark that pushing them up in post gives a lot of noise and worse. Even if not, pushing the shadows up gives me very weird colors often, it works better for black and white images.
So what do you do in these situations? Expose for the darker areas and blow highlights completely, or somewhere in between, so that the shadows can be recovered and blow as few areas as possible, or strictly avoid blown highlights?

I'm interested in how other people deal with these situations. I always read about how bad blown highlights are, but I wonder if it's really that dramatic if it's only a dull sky or parts of clouds that get blown. I'd especially like to hear from people who regularly do high quality prints, because I don't (yet), but I wonder if my files are good for it.
I'm talking about daylight photography here, when I'm at or near base ISO, low light photography is a different question.
Best,
J.

  • Decide what's important in the photo
  • Set exposure for maximum that retains the brightest highlight details you need to retain
  • Bracket and pick the best exposure to work with
  • Process to suit

I have no experience with the Sigma cameras. With the M-P typ 240, I've been able to pull up dark areas by as much as five stops with satisfactory results, color or B&W, and pull down bright areas to recover more detail as long as I didn't get to full saturation. That's really the key: don't allow full saturation in any area that you wish to have any detail. Saturation overexposure is ugly out of any camera.

G
 
Hey, Thanks for the replies, guys.
Ranchu, I'm indeed looking forward to getting access to or having space to have my own darkroom soon, for black and white work. I will probably want to keep using digital for color though.
Godfrey, is "saturation overexposure" something else than "normal" overexposure? Do you mean one channel clipping much before the others do, due to intense colors?
David, yes fill flash is a good solution in many, but not all cases. Sometimes to built-in flash is too weak or the subject stretches away from the camera, making halfway even illumination impossible.
Willie, thanks to other threads on this board I was aware of most of the things you wrote about the relationship between ISO and DR, that's why I wrote in the opening post that I'm interested in the problems that persist when shooting in daylight. If I understand you correctly, if there is no other way you will also select an exposure that might blow some highlights?
Let me try to be pose a more precise question, as I'd like to hear more opinions about what Keith wrote earlier on: Will you live with overexposing areas like sky, in which you don't need detail? How bothered are you about the transition into overexposed areas, and are you satisfied with bigger prints you have made from images with this compromise?
Best,
J.
 
Here's an example, where parts of the image got overexposed AND I had to pull up a lot from the shadows, although a little bit more would be possible and desirvble to get the girl's face a little brighter. Actually I think that one's quite ok, and I didn't have a choice. What do you think? Is it spoiled by the overexposure in the sky and the street? Could I have done it differently?
It was shot with a Pentax K20D, so the DR is not up to current standards, but I think newer cameras would have had problems in this lighting, too.
31003003.ed967027.640.jpg
 
...
Willie, ... If I understand you correctly, if there is no other way you will also select an exposure that might blow some highlights?
Let me try to be pose a more precise question, as I'd like to hear more opinions about what Keith wrote earlier on: Will you live with overexposing areas like sky, in which you don't need detail? How bothered are you about the transition into overexposed areas, and are you satisfied with bigger prints you have made from images with this compromise?
Best,
J.

If there is no other way I will either let shadows be shadows (leave them dark) or let the highlights go. It depends on the purpose of the photograph. If you look at professional interior photography in high-end magazine ads, you will see lots of blown windows. In fact you will even see window pane dividers faded away. The color in these regions is usually selectively desaturated. This makes life easy for the photographer as off-camra lighting is either not needed or uncomplicated.

Of course with raw files and optimum exposure, the shadows can be selectively pushed during post production. The degree they are pushed is an aesthetic choice and depends on the scene and the purpose of the photograph. What I noticed was – I fell into a bad habit of over pushing shadow regions just because I could. Upon viewing photos from others I realized this is common.

Skies are interesting. If you let them clip just a bit, they will take on a cyan color cast since some of the blue photon counts are completely wrong. To a point, this is trivial to fix with selective hue adjustments during post-procerssing. The challenge is to end up with a sky hue that no one would suspect to be simulated. It is easy to over reach here as well. The 1/3 auto-exposure bracketing is useful for dealing with skies. The lowest exposure is useful to simulate the unclipped sky highlights. Skies don't have details but they do have a subtle hue differences. Pixel peeping properly exposed skies will reveal a surprisingly amount of noise. This is photon (a.k.a. shot) noise and has nothing to do with the camera. Since the nature of this noise is rigorously described by physics, it can be filtered with no ill effects. Cloud detail is another matter. As the percentage of sensor site clipping increases, the subtle hue and contrast differences are lost. When the sun is near or behind interesting clouds, clipping the highlights will degrade the subtle differences in cloud region luminance.

I'm not sure what you mean by the "transition into overexposed areas". In my case high levels of highlight clipping can cause rendering issues (fringing?) at the borders of overexposed and properly exposed objects. As I acquired cameras with newer sensor assembly technologies, this problem tended to decrease. Lens rendering could play a role as well. At any rate occasionally the transition is an issue. A gradient ND filter is extremely useful in these circumstances. Otherwise one is forced to compromise.

I am am not a landscape photographer. For large prints I have not had a problem. However nothing can beat an exposure where the dynamic range of the scene does not exceed the analog dynamic range of the camera's data stream. Some images just aren't suitable as very large prints. So I would rather not compromise. I would rather use a graduated ND filter or even hand blend multiple exposures (which means no tone mapping). In both cases I use a tripod. These are common inconveniences for landscape work.

Two Ways To Clip Highlights

There are two ways to clip highlights... overexposure and over brightening.

For the first ISO is set to the camera's native or base ISO. This means there is no post-acquisition signal amplification. In this case overexposure results in the electrical charge of some of the sensor sites to exceed the maximum capacity of the pinned-diodes. Information about the true, but unknown, photon count for those sensor sites is destroyed.

For the second the sensor is actually underexposed. Increasing ISO by a stop means the light meter will advise you to decrease the shutter time by a stop. Less light reaches the sensor. All the information for each sensor site is present. However if ISO is set above the base value, the post-acqusition, analog-gain level exceeds the ADC's maximum voltage threshold. The result is the same. All the information for those sites is destroyed. This usually occurs because the light meter estimate is inappropriate. When post-acqusition brightness is achieved by digital multiplication the result is nearly identical.

Extreme sensor overexposure can cause artifacts when excessive electrical charge interferes with nearby sensor sites. This is rare... especially with newer sensor technologies. Excess ISO gain does not present this complication. Of course both the sensor and the ADC can be clipped in extreme overexposure (sun in the frame, strong lights at night, bright interior lights, bright windows).
 
Back
Top Bottom