Why I Purposely Underexpose All My Raw Images

Why I Purposely Underexpose All My Raw Images

By now it’s common knowledge that when shooting film, it’s important to not underexpose to hold on to shadows, and for digital it’s key to save the highlights. I’m going to advocate, however, that with modern digital sensors, it’s prudent to shoot underexposed all the time.

You’ll hear many people argue that there are certain situations where underexposing has its benefits, but not all the time. While it may sound counter-intuitive to the folks arguing to “get it right in the camera” all the time, there’s a method to my madness.

The first major reason to underexpose is to get a faster shutter speed. Sports photographers know that if they’re in a tough environment (for instance, a night college soccer game) then underexposing will give them the precious extra stop or two of shutter speed. The difference between 1/250 and 1/500 could mean the difference between getting a blurry mess or a sharp shot. The penalty comes on the back end, where you have to process that photo out so that it’s properly exposed, but with modern software, batch processing a large amount of files to the same exposure isn’t challenging.

But beyond the faster shutter speeds that are possible, there’s the practical benefit of protecting the highlights. Even dialing in a third of a stop or two thirds can give you the leeway to prevent a sky from blowing out, or a white shirt from blending into a white background.

I first started doing this mostly by accident more than a decade ago, when screens on digital cameras were so poor that I thought I was overexposing all of the time when I set the meter to 0. Instead, I found that things looked just fine on the computer, and I was just being fooled by chimping on the camera. But I still gained that exposure latitude, and the practice just stuck.

The original file from above in its underexposed state, shot this way to get a faster shutter speed of 1/500. It was easy to bring it back to proper exposure in Photoshop.

Of course this only applies for raw files. If you’re shooting JPG, you don’t have the editing headroom. Don’t do it!

What do you think? Is underexposing all of the time crazy? Do you do it too? Leave your thoughts in the comments below.

Wasim Ahmad's picture

Wasim Ahmad is an assistant teaching professor teaching journalism at Quinnipiac University. He's worked at newspapers in Minnesota, Florida and upstate New York, and has previously taught multimedia journalism at Stony Brook University and Syracuse University. He's also worked as a technical specialist at Canon USA for Still/Cinema EOS cameras.

Log in or register to post comments
92 Comments
Previous comments

When I worked with slides, in this kind of situation, I used to measure the exposure (spot measurement) on the white and then open 2 f-stops or 2.5 f-stops.
It's always worked.
I don't see why today, with digital, it should be different.

Film has latitude in the highlights...maybe not so much color slides but digital can be unforgiving in the highlights at plus 2 stops, depending on the camera's available dr.

2 f-stops when measuring exposure on white (so considering white as a neutral grey) is surely not out of range even for the worst digital camera;)
As I have always worked for the paper press (advertising, magazines, etc.), I have always considered the maximum latitude of the print (paper) and not the latitude of the film, which is wider. This latitude, in the best case, goes from -3 to +2.5, so I exposed in order to get the maximum detail in this range.

If you are exposing properly for snow, which is white...what other highlights other than specular light would be brighter that could blow out? Try using your camera's spot meter if you don't have a handheld one. Increase the amount of exposure from the meter reading on the snow by 2 stops to be on the safe side and see how that works. Your matrix meter in the camera is worthless here.

This is malpractice in my opinion.

The majority of data collected on the chip is in the rightmost quarter of the histogram. If you need a faster shutter speed, bump your ISO.

If you shoot with your exposure comp at -1EV all the time, and then pushing in post, you’re guaranteed to have an inferior deliverable vs exposing correctly.

I gave you a thumbs-up for correct philosophy, but your solution of bumping the EI results in the same thing; information lost. Bumping the EI results in the sensor not being fully saturated.

But one should aim to saturate the sensor as much as possible. The correct way to do this is find the part of the highlight in which you need detail, and bring that highlight close to saturation. Any brighter highlight will be blown, and that is fine.

We don't always need detail in the sky, the snow, the sand, et al. This idea of always underexposing so as to save the highlight is a bad idea, but, if one needs a faster exposure, by all means, decrease the exposure time and raise the shadows in development. (That is all “bumping the ISO” does).

It's all relative to end use. Web, print, cmyk, size...

I deal with very fast cars that I shoot at night with poor ambient light and strobes. There are things to let go and things to go for. I want some of the scene background illuminated but the cars are the most important. One car can be totally black and the next totally white. To me it's about a basic set up and from there playing with light if I get the time to change the strobes output, ISO and exposure variations. That's my real world situation. I can't control everything all the time, but my aim is press print, so I capture what I know will come close to what I want to do in post. Proper exposure is not my priority, what I want to expose for always is.

That's only true if you want to raise the shadows. But for night sports often one wants to darken the shadows to prevent background distractions from competing with the subjects. In order to prevent blowing out the details in the white jerseys on the field, it's necessary to shoot at a lower exposure than the light meter, which is trying to average everything to gray, will recommend.

Sometimes you really do want the shadows to be pure black.

Everything you have to deliver to a client has to be processed, and so I think the inferior deliverable would be the picture with the blown out highlights, rather than a photo that needed shadows raised by a stop with no loss of detail.

I'm not talking about blowing out highlights, I'm talking about exposing to the right to get optimal signal like what Karim explained.

I mostly underexpose; even JPEG give me enough latitude for my landscapes.

Guys, underexposing so you can keep the iso low or the shutter speed high is nonsense. When you bring it back up in post that's just like increasing the iso. Or worse, depending on the camera and where the dual gain kicks in. The image quality will suffer. The only valid reason is to protect the highlights when you're using low isos in plenty of light.

Except that it lets you preserve the highlights on the white jerseys on the field at the expense of leaving the dark parking lot outside the end of the stadium, or the tree line over the stands, pure black.

It's not really under exposure if you're barely not blowing out the details on the white jerseys. It's just that your camera, which thinks everything should be averaged in the middle, thinks you are underexposing.

Yes, from a pure exposure standpoint Rk K, you are correct, but I'm talking about getting that faster shutter speed to stop action, which is sometimes necessary.

You get faster shutter speed by increasing the iso. You only underexpose if you need to preserve the highlights in a contrasty scene.

Depends on the camera. On some cameras shooting the higher ISO is a better call (older Canons) since bringing up the shadows produces a lot of noise in the image. On my Nikons, I've found I get cleaner results by bringing the photos up a stop or two after the fact as I describe in this post. YMMV

That's not only impossible, but the complete opposite of what every lab test says. Unless you have some scientific testing of your own that you're hiding it's a stupid, irresponsible advice to give.
You can also loose the advantages of dual gain with this. A lot of cameras have dual gain now.

Lab tests can say whatever they want. Real world experience says another.

Stupidest saying there is without context.

At least show some evidence before misleading people. This is madness!

I bet you haven't even tested this recently, just a bad practice you picked up a long time ago.

Image quality and noise are easily quantifiable, objectively measurable. It's not going to change from a lab test to real life shooting, like autofocus or ergonomy would. But in a lab you can actually find out the truth.

Samsung's Galaxy Fold phone worked perfectly folding thousands of times in a lab, but then five minutes into the hands of reviewers in the real world revealed that the lab tests didn't mean much. It's much like DxOMark's sensor tests - interesting to read about, but in the real world, things are much different. It's why we can complain all we want about the dynamic range of say, a 5D Mark III compared to a Nikon D750, but there are plenty of both cameras to go around.

Completely irrelevant and off topic whataboutism. I asked you to show me an example where raising the iso is noisier than raising the exposure in post. You can't, obviously, because it's a ridiculous, completely unfounded claim.

The sensor doesn't know or care if you're shooting in a studio or at a football field. Nothing changes from that perspective. That's why the lab test is conclusive for this.

And yes, canon's ancient sensor tech is starting to cause problems for them, just look at the sales of their new mirrorless line.

I learned a long time ago that it was better to push the histogram slightly to the right (overexpose), mainly because 75% of the bit information in a RAW image was reserved for the top 25% of the exposure, the highlights. By pushing the histogram to the right, the overexposed areas could be rescued with the post process exposure control, with more bits dedicated to the color depth than information recorded in the mid range or shadows. The darker areas could be pushed further into the shadows without crushing, and by not having to pull shadows up, there was less noise to contend with.

Was this accurate? At the time, about 10 years ago, it sure made sense. I suspect that processing power makes the point moot these days, but I still fall back on it once in a while if I think the image warrants it.

I believe we are in better shape now for doing whatever we need to do in order to optimize our images, but I will still defer whenever possible to avoiding noise if I know I can rescue highlights completely.

i use Highlight-Weighted Metering Mode in some occasions,.

In which case you are getting about the same result at "zero" exposure compensation as someone else using Evaluative/Matrix metering is getting with -1 EC when shooting field sports under lights at night.

Oh, is it so? So what you saying is that camera matrix metering has tendency to read light one stop lighter? Yeah... it seems some cameras has different tendencies. I was talking about exposing with meter...

I'm a firm believer in exposing properly and not correcting my mistakes unless it's absolutely necessary. By underexposing, you are losing everything your camera can deliver.

Wrong, it's NOT a mistake, it's intentional with a purpose.

Doing something intentionally doesn't mean it isn't a mistake. It's a willful mistake.

You are reducing the dynamic range of the image, you are reducing the signal to noise ratio of image. You are reducing the color response of the image. All because you are afraid of or aren't able to prevent over exposing.

Mistake is as mistake does.

No mistake, no one is afraid, just photography as it has always been.

It's not really underexposure if you still have to use highlight recovery to keep from blowing out the details in the white jerseys of your subjects. It's more that the camera's meter is being fooled by all the dark areas in the background, like the parking lot outside the end of the stadium, or the stands that are lit 3-4 stops dimmer than the field of play is. The camera thinks you're underexposing, but you really aren't. You *want* those dark shadows in the parking lot or in the trees behind the top of the stands to be pure black. You don't care if you lose details in those areas, and you're not going to raise those shadows in post. If anything, you're going to crush them to pure black.

Newsflash 2019, 2018, 2017, 2016, etc.... In many situations, underexposing IS exposing correctly. And, I don't mean -5 stops for no reason. Like many things, you have to find that balance of what your intent with the image and the abilities of your camera.

To bypass shutter speed challenges for moving subjects, why not shoot video and pull out frames in post-capture processing, if we're gonna do anything in post?

I actually like the darker original image, because there's too much visual "noise" behind and surrounding the presumed main subject, so "low key" is a useful choice, and available as an intelligent camera capture program.

Otherwise, it seems like I'd be trying to compensate for not knowing how to buy a camera I trust to work the way I work.

I know how to select photographic engineers first, then buy the camera they build, and they've never let me down.

If I have to use manual settings for the normal routine situations I photograph in near proximity in real life, then I bought the wrong camera.

I suppose that the opening example would challenge me to I know how to compensate for my camera's limits, and if don't mind extra work in post capture, then at least I'd become intimately familiar with my images by scrutinizing them up close and personal, one by one as I make adjustments later.

If that's a significant contributor to what I enjoy, then I'm happy.

Though, I've never manually shifted my modern automatic transmission car, but again, I chose the driving engineers first, then I bought their car -- for me, offloading those arcane calculations frees me to have fun, so I enjoy driving and getting somewhere.

In a like manner to putting my car's transmission in [ D ] for [ Drive ] and then driving, I set my camera on [ P ] for [ Professional ] and shoot away all day without chimping or post processing ...
.
.

Because video frames are lower resolution and show much more motion blur at typical video shutter speeds used to keep the video from looking "choppy."

Lower resolution than the target presentation?

Again, this begs for using appropriate gear.

But I suppose that if using compromised gear and playing with it gets a satisfied customer, then that's really the title of the article, in this case:

"... How to compensate for an unintelligent unsophisticated inappropriate camera exposure system ...":

Thanks.
.

In some situations, you can make a decent usable print from a 4K video grab (gets you about 8MP of still image), but for the reasons Michael Clark listed, this would not be practical in video. Not a matter of compromised gear, just a reality of video capture vs. still capture. Two different animals.

Underexposing by 1/3 of a stop has become my default setting on the 5Div, however I never felt the need for it on previous Canon nor on any of the Sony A7 series bodies.

This technique isn't really underexposure in the classic sense. It's more of meter calibration (if using semi-automatic exposure modes) or proper manual exposure even though the camera's meter thinks you are underexposing. In the context of night sports, there are a lot of areas in the background you *want* to stay black. You're not going to try and push those deep shadows in post. You might even crush them further!

I set exposure so that the white jerseys (one of the two teams is almost always wearing white jerseys) are only slightly *overexposed* in the jpeg preview. In post I then pull back the highlights more than I push the mids by pushing overall exposure and also crush everything within two stops of pure black to pure black. When exposure is set to do that, the camera will typically protest that I'm about one stop underexposed. But It's not really underexposed if I'm going to raise the mids but lower the highlights *and* the shadows! It's just that the camera does not understand that those dark shadows in the background are *supposed* to be dark.

Some people don't mind if the subject is perfectly exposed and the background is overexposed as long as the falloff is very clean. Sometimes people slide the highlight slider too far when it's better off too just leave it the way it is. I found that I was often happier with a picture when I don't touch it as much in post production. But everyone is different its all about style.

Since not blowing out the highlights is critical, especially in a dark scene with important white elements, I'll typically use AUTO-ISO in manual mode and set the body to 1/1000 th second and use matrix metering with highlight priority. Set the camera and forget it. This will work in most situations and your images may be slightly under
exposed to save the highlights.

I think the general advice would be get the exposure right, and than adjust accordingly... depending on your lens (some lenses are so contrasty that you loose details in shadows or highlights much faster), or situation. Shooting under harsh sunlight - there maybe some peace of heart in underexposing... but in reality, even if you overexpose by 1 stop, usually it comes back fine in Capture One or any other software. If you feel like you need faster shutter speeds... use higher iso... youll be mostly safe till 540 zone.

I see a lot of amateurs who underexpose images, but if you working in editorial - unless there are some artistic thought behind it - you expose everything properly.

Under-exposing is a standard practice for me. I had photographed sports for years with the top Canon Sports camera 1DX series ... and even with the best professional f/2.8 lenses, it is nearly impossible to get fast enough exposures (with no smear in fast movements) with ISO levels below 800 and a (0) exposure compensation setting at the lighting levels for nearly all stadiums and indoor sports venues worldwide. As a general practice, -1 full stop under in RAW is easily recovered. The results speak for themself at least in the case of Canon 1Dx and 1Dx MkII. What I dont know and would be interested in an academic study of how many stops underexposure and software recovery of RAW images is reasonable before there is a loss of image quality equal to simply jacking the ISO. Emperically - i found the limit around -2.25 stops

I've read a lot of the comments below and one thing stands out. It has to do with the base exposure reading. Are you underexposing from an "in camera matrix" meter reading of a scene with a lot brightness above mid-tone? If so, you have really buried your shadows as the original meter reading will provide an underexposed image to begin with. There is a way around this...a way that will provide you with proper exposure readings that already account for highlight protection. It just seems as though digital shooters...most of us anyway, don't want to invest the money in or time to learn handheld metering. We go out and buy a camera and lenses for thousands of dollars. Equipment that is designed to provide the best file quality when that file is properly exposed. And, then we fool around with looking at a histogram which tells us absolutely nothing about an exposure based on mid-tone of the incident light and then we depend on the camera's dynamic range as a rescue tool. I hear: "...meters are technical and interfere with creativity." (Do measuring devices interfere with a top of the line cabinet makers artistry?) , I hear too: "Metering takes to much time." (So your thing is what? Competitive speed photography?) This isn't an all or nothing deal either. I don't meter casual shots of the grandkids birthday parities. But I meter every shot of the grandkids in my studio. And, I measure every shot in ambient light that could be important to me for capturing beauty, telling a story or for having a memory about how I felt while taking the shot.

So, how does this work out? Well, if there are bright objects in my image, my histogram shows exposure to the right, but safely so with the highlights protected. If there are black objects, the histogram is pegged to the left, but shadows which are not totally black are protected. In other words, the tonality in my files accurately reflect the tonality in my image. Luminance, saturation and tonal gradation are all natural.

But, and seriously, I'm not trying to start a meter/no meter war. There is no right or wrong. Just methodology. I'm just like the rest of you explaining to this thread about how I go about exposing and protecting highlights.