0
\$\begingroup\$

Sometimes I have shots that are not well exposed (I shoot in RAW) and it's really dark. I would selectively brighten up the shadow area.

Usually it's fine if the area is not too dark. But if it's really dark (imagine an incorrectly exposed image, exposed for the window and the person inside has a silhouette) the color becomes weird when brightened up. A person skin would look green. And it's really difficult to correct. In fact, using the white balance tool doesn't even fix it. I have to manually tweek the color curves (red green blue) and it's impossible to get it back to looking like normal skin.

Why does this happen and how do I fix it?

\$\endgroup\$
2
  • \$\begingroup\$ Can you post an example? \$\endgroup\$
    – AJ Henderson
    Jun 5, 2013 at 16:17
  • \$\begingroup\$ "How do I fix it?" Expose your shots properly in the first place - RAW isn't a magic bullet; while you get more latitude to fix mistakes, you don't get an infinite amount. It sounds to me like you're trying to pull data out of an area where there isn't really any. \$\endgroup\$
    – Philip Kendall
    Jun 5, 2013 at 16:25

3 Answers 3

2
\$\begingroup\$

This is normal. When you brighten something, you are amplifying pixel values. This causes any randomness to by multiplied. Noise has a greater S/N ratio in dark areas which is part of that randomness. When you make things liker, the noise becomes lighter and when its effect on color gets multiplied too.

Depending on your camera, noise may be more present in one channel. The same effect is visible when you shoot at high ISO since the camera is doing something similar. Because of this, you often will see images turn greening or yellowish or even bluish on some models.

This cannot be corrected by white-balance since this is not a change in color temperature which has a defined progression from cool to warm. It is possible to correct it with tint but the most direct way is to use color curves.

\$\endgroup\$
2
\$\begingroup\$

Some of your channels dont make it through the noise floor. It is better to expose to the "right" - meaning that you seem to be overexposing - the jpeg looks all white in the light areas, but then you pull down the highlights in the raw editor, and voila! you got your perfect image. When this doesnt work, you need to resort to HDR imaging. Just dont overdo the HDR processing. It is possible to get a decent natural looking result out of it. Even better is to avoid shooting against the bright window in the first place, walk around, get a better angle of your subject, just just snap a shot from where you are and leave it at that.

\$\endgroup\$
0
\$\begingroup\$

Digital cameras work by sampling red, green and blue light. If the light has a tint to it or the image is severely underexposed, it is possible that there is either not enough light registering on a particular color or there is not actually enough of a particular color present to get a reading.

Since normally we expect a range of tones, if only one or two values of tone are sampled over the entire image, then no amount of expansion is going to produce a natural looking image since the information needed is simply not present. It is the same problem as trying to colorize a black and white photo. If their wasn't enough color depth on a particular color, it can't blend back.

You see the same problem with things like underwater photography or shooting with certain types of LED light where certain colors of light are massively missing from the image, just in the case of severe underexposure, it is too small of a range of values instead of not being present. In your case, there may have happened to be more green light, so it got more exposure there, but wasn't able to get enough red or blue information.

This isn't necessarily the only possibility though, so sample photos would be helpful for verifying the issue.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.