Noise reduction: Overexpose, underdevelop

Have you heard this formula? If you first camera was Kodak Instamatic you probably have. If you first camera was digital you probably haven't.

Overexpose, underdevelop is an older (though still widely used in practice) way of shooting. In short it means that one should give more exposure to film then it should be for given ISO and luminance of the scene. What the overexposure gives us is detail preservation in shadows. Isn't it great?! Well, not exactly. Because overexposure blows out highlights. To preserve highlights one needs to underdevelop. Thus this approach allows better detail preservation/grain reduction in shadows. You may dig Internet for better details why it happens. In shot, the approach uses non-linearity of sensitivity curve of film and developing process.

Can it be used with digital cameras?

No. It should be used with digital cameras, particularly if you shoot RAW, although JPEG shooter will also benefit. In the digital world, it sometimes called exposing to the right (moving histogram to the right).

In the digital world, one needs such an exposure that a histogram of the captured image just started clipping on the right side. Highlight warning mode is at a great help here to ensure that nothing significant is clipped. This is an overexposure step.

As well as in the film world, the overexposed image need to be underdeveloped in a camera raw processing software to bring washed out image to an appearance of normally exposed image. Generally, it means you need to move an exposure slider in an RAW processing application toward the left.

And the result? Yes, it will be exactly the same - detail preservation in shadows and noise reduction. Yes, noise reduction. And yes, not only in shadows but also on saturated colors, like blue skies and red sunsets. Yes, you're absolutely right it will be full frame noise reduction, just because during the capture and development noisy bits of the image are suppressed.

And yes, it's absolutely free of charge and nobody yet patented it.

Unfortunately my images which I could use as samples were rejected due to excessive noise. But fortunately it's not very difficult to see the advantages of the technique. Just go out and shoot blue skies or if you're missing ones then look for deeper shadows and smooth darker tones.

This two images were overexposed during capture and underdeveloped during RAW processing stage. It helps reduce noise on darker part of the tea cup (right image). I did not have much of latitude due to specifics of light so I overexpose about 2/3 stop. The left image was shot with about 2 stop overexposure since lighting was even and contrast was low. Though blades of windturbines lost their details but it was not that important for this particular shot. No post-processing except exposure adjustment, sharpening was applied for both images.

How much to overexpose? As you may see above - it depends on particular lighting conditions and also on your camera. So turn on you highlight warning and histogram and experiment. On Canon 350D I have about 1/2 so called headroom. That's mean that if an image histogram has reached the right side, I still can add about 1/2 stop without loosing details in highlights. The headroom vary from camera to camera and also depends on light source. Typically cameras optimized of day light and departure from this type of light reduces camera's headroom.

JPEG shooters can use the same technique but they should be more conservative on the overexposure side, since they don't have much of headroom.

Photo credits: Actionwatcher.

Your post must be written in English

September 20, 2007


Thank you. I'm glad to be at help

September 20, 2007


Interesting read. Thank you.

September 19, 2007


Excellent questions.

Regarding combining differently exposed images:

The shot answer will "No". If light allows overexposure one overexposed image will most likely take care of both noise and shadow details with much less work on your side.

The long answer will be "Yes". Is it funny :). If you would like to achieve higher level of noise reduction or there is no much space to the right to overexpose then two frames would helpful. Moreover, there is a lot of cases when light is too harsh thus shadows are blocking (too dark) and highlights tends to blow out. In such cases combination of two or more differently exposed images will solve the problem of shadows and highlight detail preservation. This technique can be called either HDR or extended dynamic range. Photoshop CS2 and CS3 and also number of other software applications will be at great help. Unfortunately, if the dynamic range is too high then the final image may look unnatural and generally will require a lot of work and experience to make it good or great. And there is a lot of examples of successful work. If you combine only two frames you may use Photoshop's PhotoMerge in File->Automate with following masking work.

About Pushing.

We, digital shooters, already have it. When we increase ISO on our cameras it is what can be called digital pushing. Thus in exchange of higher noise we are getting greater sensitivity (exactly like for slide/film). Pushing digital images in RAW converters would give little or no benefits if compared to increasing ISO on a camera.

Thank you for the questions.

September 19, 2007


Having shot some difficult nature landscape images a couple weeks ago, I'm beginning to wonder if using a combination of underexposing and overexposing an image, and then combining the two digitally, would get the desired effect..??? I may have to play with that one..

September 19, 2007


Another term not heard very often any more is "pushing film." If you had 200 iso film, but the conditions called for, say, 400 iso film, you could "push" the film to perform closer to 400. You would do this by setting the camera to 400 iso settings, but loaded with 200 iso, and shoot like normal. In the lab, you had to develop the negatives as 400 iso, not 200 iso. For those with their own labs, it was a bit easier. To rely on a photo lab to do so didn't always get the results you were expecting, as the request was not always read, or they refused to do so for one reason or another.

Related image searches
Noise related image searches