Dealing with noise in HDR photos

Article by Wojciech Toman

In this article I will focus on dealing with noise in HDR photos. If you would like to learn more about HDR photography and about my HDR workflow read my comprehensive HDR tutorial describing each step of creating HDR photos.

Noise is an inevitable part of digital photography and it is related to the modern sensor design (and also technical limitations), and results in either some ugly monochromatic grain (luma noise aka luminance noise) or even worse, colourful dots or patches (chroma noise aka colour noise). Without going into technical details, noise becomes stronger with higher ISOs and longer exposures. At certain ISOs it’s so high that the photos are virtually unusable. Now, how does this relate to HDR photography? The thing with local tone-mapping (like one in Lightroom or Details Enhancer in Photomatix Pro) is that it has a very bad feature of amplifying noise (because it enhances local details and noise can be consider as details in fact). It means that noise in the tone-mapped image might be even higher than in source photos.

Hence noise becomes an even bigger problem in HDR photography than it is with “regular” photography. Reducing noise is always essential but with HDR photography it’s even more important. Just take a look at this highly exaggerated example. The noisy part is completely ruined. You don’t want your photo to look this way, right?

So here are some tips on how to reduce noise in your tone-mapped photos. Hopefully thanks to them your image will be as clear as the bottom right part of the photo above:

1. Use low ISO

Shoot at your camera’s native ISO whenever possible. This doesn’t mean lowest ISO available! For instance on Canon 5D MK II (and many Canon cameras in general) you will get lowest noise not at ISO 50, but at ISO 100. For many cameras you have option to set ISO to 50 or 100 but it may turn out that native ISO for this model is in fact either 160 or even 200. Google your camera to find out this value.

2. Watch out for blue channel noise

Watch out for the noise in the blue sky and in the blue water areas as there is generally more noise in this areas. Take a look at the image below which is full of blue color:

Digital camera sensors are by design rather insensitive to blue light (they are about 50% more sensitive to red and green) so in order to “expose” the blue channel correctly it has to be slightly amplified. If you’re familiar with the signal to noise ratio term then you know that there is more noise in darker areas than in bright areas. If blue is dark and it is brightened, noise will become more apparent.

What’s more, the Bayer filter used on camera’ sensor to produce colour data instead of a monochromatic image results in twice as many green pixels as red or blue ones. Why is it this way? It’s simple – human eye is most sensitive to green, then red and blue parts of the light spectrum. Take a look at the image below depicting Bayer filter. As you can see there is twice as many green pixels as blue ones (you can count if you don’t believe me):

Now, this gets even worse when you use a polarizing filter which darkens the blues in the sky because in post-processing you would probably need to brighten it a little bit. What’s even worse the sky doesn’t contain any texture. As tone-mapping enhances details, noise gets much more apparent there than in texture-rich regions of the image.

For all these reasons it is a good idea to reduce noise in the sky selectively. If you were to use the same values for the rest of the scene it will get blurred decreasing clarity and removing fine details. There are plenty of ways to do this. You can either use Blur tool from Tools palette or select sky (eg. with Magic Wand tool) and use noise reduction filter of your choice on this part.

3. Reduce noise prior to creating HDR

Reduce noise on your images prior to merging them to HDR. The reason should be obvious – if there is less noise in the HDR image, it will get amplified much less during tone-mapping phase. Also it is a good idea to treat each file individually as different photos might have different level of noise in them.

Photomatix Pro reduces noise before merging photos to HDR if you select below option in preprocessing window:

4. Use dedicated noise reduction software

Use dedicated noise reduction software. Many applications like Lightroom or Photoshop have noise reduction features built-in. However, these noise reduction algorithms are pretty simple compared to the ones implemented in software like Topaz Denoise or Noise Ninja, eg. Topaz Denoise offers you controls to reduce noise separately in shadows, highlights and also in red and blue channels.

Below you fill find a screenshot from Topaz Denoise I use to denoise my photos:

Usually I start with selecting a preset on the left. I try to find first one which removes all noise in the image. Usually it blurs details a bit too much so I choose preset just before it and fine-tune it until I get desired level of noise reduction.

5. Reduce noise at 100% magnification

Similarly to sharpening you should reduce noise while viewing your image at 100% magnification. Otherwise the image is downsampled what might give you false impression that noise is much lower than it really is. It’s because downsampling results in interpolating neighbouring pixels is what has the effect of blurring tiny details and noise.

6. Cover whole dynamic range of the scene

Make sure to cover whole dynamic range of the scene – if you don’t properly expose darkest shadows the noise will get amplified in them. By properly exposing the shadows I mean that the brightest photo in your sequence should have darkest shadows in mid-tones part of the histogram.

7. Use lower exposure value spacing

Shooting at lower EV spacing (eg. 1 EV or lower instead of 2 or 3 EV) will also result in lower noise in final tone-mapped image. This of course means that you will need more photos than with 2 EV spacing. Many Nikon cameras allow shooting more than 3 bracketed photos so it’s not a problem for Nikon owners. But what about Canonians who are most of the time limited to 3 bracketed photos? You might consider installing Magic Lantern on your camera which will let you use any number of photos.

8. Use loupe in Photomatix

When tone-mapping in Photomatix, use the Loupe window to assess level of noise. As it offers a preview of 100% crop, this window is also useful for assessing details enhancement so it’s always a good idea to use it. To open it just click left mouse button anywhere in your image.

9. Make sure to reduce noise enhancements in Photomatix

In Photomatix there are two settings which can help you greatly reduce noise in tone-mapped image: Micro-Smoothing and Smooth Highlights. By increasing them you can lower noise level in tone-mapped image. When you’re changing them, make sure to have Loupe window opened (take a look at point 7).

10. Use other processing methods

If you can’t get good results with local tone-mapping, try using either global tone-mapping or Exposure Fusion instead. The latter has a nice effect of reducing noise.

With above tips you will greatly minimize noise in your HDR images, making them more pleasurable to watch even at 100% magnification. Please note, however, that there is no method of removing noise completely because noise in digital photography is unfortunately as inevitable as death and taxes.

Article written by Wojciech Toman

Facebook – https://www.facebook.com/wojciech.toman

Blog – http://hdrphotographer.blogspot.com/

  • http://thefella.com/ TheFella

    Great article with lots of very useful info!

  • http://www.outofchicago.com/ Chris Smith

    Nice article. This is something I probably don’t put enough work into. Usually I just drag the Lightroom slider to 30 and move on! That first image is great! I love it.

    • http://www.facebook.com/wojciech.toman Wojciech Toman

      Thanks Chris. From my experience dragging noise slider in Lightroom to 30 greatly decreases image quality (especially in the fine details) so I recommend using some other tool :)

  • tech.blog3

    Great article

  • Adithya

    Really helpful! Thank you for the very informative article.