05 Choose an area
Now click the sampling tool in the tools panel (circled) and draw out a marquee on the image in an area where you want to measure the noise (also circled).
06 Reduce noise
Once you’ve got your noise profile sorted out, click the Reduce button (circled) in the tools panel to apply your noise reduction settings.
This is often easier if you the zoom button on the top toolbar to view the image at 100% magnification, and click the split-screen view button (both are circled here).
The noise reduction controls themselves are really simple. Noise comes in two types: luminance noise (Dfine calls it ‘Contrast’ noise) and colour noise, and there is a slider for each.
Colour noise is easiest to deal with. The default value is 100%, but you can usually increase this without any obvious harm to the image. There might not be much obvious benefit either, though, because most modern cameras control colour noise really well, so you probably won’t have any in the first place.
Luminance/contrast noise is the tricky one. This is where most of your image noise will come from, and while you can reduce luminance noise you also start to lose the textural detail in the image, which can give you an over-smoothed, ‘glassy’ look.
I usually fine Dfine’s default value of 100% too high, so I reduce it slightly to recover some of the lost textural detail. As you do this, the noise starts to reappear, but it’s not hard to find a balance that gives you the best overall quality.
07 Control points
Dfine 2 does have one more trick up its sleeve. Like the other Google Nik plug-ins, it has control points for localising the effect. In this case, I can place a ‘minus’ control point over the clock face of Big Ben to remove the noise reduction from that area – this gives me back some crisp fine detail where the image most needs it.
I don’t think the noise reduction in Dfine 2 is necessarily better than Lightroom’s, for example, or DxO’s. The point is, however, that noise reduction process brings drawbacks – you always lose textural detail along with the noise. This means you have to make very careful compromises on an image-by-image basis, or even from one area of the image to another, and this is where Dfine 2 really scores over in-built noise-reduction tools.