In my recent comparison between DxO’s new DeepPRIME vs Topaz DeNoise AI, it was clear that DeNoise AI doesn’t do as well with smooth backgrounds.
You can read the DeepPrime vs DeNoise AI comparison here
The Problem
As you can see in the two images below, the background of the DeNoise AI image has a small mottled effect that makes it less smooth than the DeepPrime image:
DeNoise AI (first) shows a subtle mottled background compared to DeepPrime (second)
A closer look better reveals the problem. DeNoise AI has definitely rendered the background in a less pleasing way than DeepPrime. Even in a printed image the difference is fairly easy to spot.
DeNoise AI
DxO DeepPRIME
The Solution
Luckily, it is quite quick and easy to fix this issue if you have access to a pixel editor such as PhotoShop or Affinity Photo. I will be using PhotoShop for this tutorial, but the same steps will apply in all editors.
Follow your normal routine to create your image and apply Topaz DeNoise AI. Open the image in PhotoShop
Duplicate the background layer by right clicking on the layer and choosing ‘Duplicate Layer…’ or just press ctrl+j (cmd+j on mac).
Duplicate the background layer
- With the new layer selected, Go to the ‘Filter’ menu and choose ‘Blur -> Gaussian Blur…’
- Now apply a slight blur to the layer. The aim is to keep things subtle and apply just enough radius to make the mottled background disappear. Don’t worry about the rest of the image being blurry – we will deal with that next.
- Now the background areas of the image should look smooth. Obviously the rest of the image looks pretty bad, so now we will fix that. Create a layer mask on the blurry layer.
- Next, choose the brush tool and select black as the brush colour. Any area you paint black on the mask will reveal the non-blurry photo from the layer below
- Finally, paint over the areas of the photo that aren’t part of the background. The great news is that you don’t have to be too precise with your brushing around the edges, in fact you will need to brush over the edges to remove the blur.
Paint over the non-background areas in the image.
Now you should be looking at the finished product. The background should be extremely smooth and the other parts of your photo will look as they should.
Tips
If you’re not familiar with using layer masks and brushes in PhotoShop, the below tips may help
If you make a mistake when brushing and accidently brush over an area that you shouldn’t have, all you need to do is change the colour of the brush to white instead of black. A white brush will paint in the top layer, a black brush will remove the layer.
If a black or white line appears when you use the paintbrush, it means that you are painting on the image instead of the mask. Make sure the mask is selected in the layers palette, not the image. The mask should have a border around it.
Notice the small white frame around the layer mask showing it has been selected.
If you’re not sure which parts you’ve painted over, you can take a quick look at the layer mask thumbnail as a guide (as per the screenshot above). You can also press the backslash key “” to show the painted area as red. Press backslash again to turn the mask view off.
The red areas show where you have painted on the mask. It matches the black areas in the layer mask thumbnail
Why Not Just Use PhotoLab?
All of the above seems like quite a bit of effort to go to when you could just use PhotoLab 4 to get a smooth background from the get go.
Well… I do.
For the most part, I now use PhotoLab 4 as my first choice when it comes to noisy images. However, PhotoLab does have some limitations in some situations
It will only work on RAW images from a supported camera (list here)
PhotoLab 4 (US$199) is much more expensive than Denoise AI ($US79.99). It is worth it as a RAW converter, but maybe not if all you need is noise reduction.
So, if you already own Topaz DeNoise AI and don’t want to buy PhotoLab 4 Elite just for noise reduction, or you work with JPEGS or a camera that isn’t supported – then you can get great results with the above technique.
Once you do it a few times, you will be surprised how quickly you can apply the steps mentioned above.