Excessive HDR Noise in the Single Image Process

Single image HDR processing is pretty cool.  It can give new life to some old photos that were taken at a time when HDR was never even a consideration for your images.  The one problem with single image processing is that you can get a lot of noise in your image that was never visible in the original RAW file.  Why is that?  Well that’s because the HDR process is trying to expand the tonal ranges that exist in your single image and, without a variation in exposures for the process to draw from, it has to try and expand the existing data.  This is especially troublesome in shadow and dark areas where noise that is not readily apparent is amplified.  This is especially true when you bump the detail in the microcontrast sliders in Photomatix tonemapping.  So here is an image that was originally shot in RAW.

Here is a 100% enlargement of part of the original.

Now here is the tonemapped version of the file that was converted as a single image and then tonemapped in Photomatix.

Let’s look at this image at 100% to see the elevated noise levels.

You can see in the sky area that the noise levels have risen to a very high level as a result of the HDR and tonemapping process.  So is there any way to combat this?  Well, there are probably a few ways but the way that I deal with it is to run a good noise reduction program on the image such as Nik Define 2.0.  This Photoshop plug-in is very powerful stuff and it also places the noise-reduced version into your document as a separate layer that can then be masked to put detail back into the image as needed.

Here is the same image after running it through Define and giving it a few more tweaks in Lightroom.

And the same image at 100%.

You can see that the noise levels in the sky have been dramatically reduced.  Now that you know about the noise issue from single image processing, you can combat it in your camera by shooting enough exposures to get good detail in your shadows, which will reduce this noise substantially in your tonemapping.

Switch to our mobile site