On the weekend I attended a wedding and took my camera along. I was not the official photographer -- just a guest.

My camera is a Canon 5D Mark III, and I was using an EF 50mm f/1.4 USM lens. At the evening reception, which was a dark room with disco lights etc, I was going round taking lots of photographs of all the party people and in order that I didn't have to worry about my camera settings, I shot in Aperture Priority mode, with auto-ISO, and f/number of mostly around f/2 - f/4. I used AI Servo mode so that it would track people dancing etc. Of course, I also shot in RAW. It is a testament to the 5D Mark III that it was able to find focus in such challenging conditions!!

However in my stupidity, I didn't change the maximum auto-ISO setting, and most of the shots were taken at a whopping ISO 25,600 -- the camera seeming to prefer to increase the ISO rather than slow the shutter speed to get an exposure. The shutter speed in most shots is probably faster than it really needed to be - around 1/125-1/250th... For a 50mm lens I probably only really needed 1/80th-1/125th to freeze the action?

This has resulted in some incredibly noisy and grainy shots, lacking in the detail I'm used to. Note that I don't think it's a focus issue - the focus is fine, and the shots aren't blurry either. Just noisy.

I imported all the photos into Lightroom 4, and have been playing with the Noise Reduction and Sharpening sliders, but in order to really get the noise levels to a normal or acceptable level, I've had to push that NR slider all the way up to 80 (choke!!). Normally I'd only apply at most about 40, even in a low-light photo! This much NR has meant that the detail in peoples faces is overly softened, and skin tones are looking almost plasticky.

So my question is, what would you say is the best way to "recover" these photos and balance noise reduction against sharpening, or to reduce the noise in another way? I do also have Photoshop Elements 11 at my disposal, though I'm not the best with it. However, if there is another way in which I could use that to achieve the same thing, I'd also appreciate some pointers.


3 Answers 3


When I have extremely noisy images, I do two things:

  • Use a 3rd party noise reduction plugin - in my case I use Topaz DeNoise - it, and others, have free trials - so you could give them a try if you want to experiment.

  • These denoise plugins have sliders that will reduce noise, which softens the image, but you also have control over detail (you can decrease noise but retain detail on edges) and you can also control separately whether noise reduction is applied to highlights, midtones and shadows. So I might first limit the noise reduction to the shadow areas.

  • I do this in several layers, one for dark background with no detail, where I push the noise reduction all the way - not worried about detail. Another layer or three for other areas which have detail I want to preserve.

    • Using layers I can then mask in different parts of different layers.

In your case, I would apply as much noise reduction as I could in LR4, not to the point that you lose detail. Then export to Elements, try out a plugin like DeNoise or Noise Ninja, and work in layers.

  • \$\begingroup\$ FYI - Noise Ninja is now Photo Ninja and is a full raw conversion application that can stand-alone. I prefer Topaz, myself, but Ninja is quite good. \$\endgroup\$
    – Joanne C
    Mar 4, 2013 at 22:27

One thing you can do is to convert to black and white and use the noise for effect.

I believe that sensors (and I confess I forget why) have twice as many green sensors as as red or blue, when converting to B*W only use the green channel and you when you reduce the noise using your favourites editing software (I use light-room and it is very effective) you should get better results than using all channels.

  • \$\begingroup\$ The reason is that the human eye is most sensitive to green-yellow, less sensitive to red and much less sensitive to blue and violet - think of how much darker full-saturation blue seems than full-saturation yellow. Cameras get away with storing less blue and red information due to this. \$\endgroup\$
    – JohannesD
    Mar 4, 2013 at 22:06

Reducing noise by automatic means is a tradeoff with losing some detail. That's just math. It seems like your pictures have such high noise that there is no place along the noise versus detail tradeoff you consider acceptable. When the signal to noise ratio isn't there in the first place, you can't really reliably recover it.

However, signal to noise ratio is at a particular resolution. Put another way, you can trade off resolution with signal to noise ratio. If you can filter down the pictures to half or less of their size in each dimension, then you will see substantial improvement. If these smaller size pictures are acceptable, then you're done and nobody needs to know how noisy the full size versions are.

If you really need the full size images without the noise, then you're screwed. Don't let the camera pick stuff for you next time.


Your Answer

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

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