Dfine is the Google Nik Collection’s noise reduction tool. It’s designed to improve high ISO, high noise images rather than to create any particular image effect. It opens with a split screen view of the image at 100% magnification with the untreated image on the left and the treated image on the right (you can change the view mode). There are just two stages to the noise reduction process: Measure and Reduce.
The Measure step checks the noise properties of the image, and the default is for Dfine to do this automatically, selecting one or more areas of the image for characteristic noise patterns. These are indicated by rectangular marquees, and if you think you can make a better job of identifying key areas of noise you can take over, switch to manual mode and use the marquee tool to define areas yourself.
Whether the noise measurement has been done automatically or manually, Dfine will apply default noise reduction settings, and you can see the effect in the ‘after’ half of the split screen display.
The default noise reduction is pretty strong. Dfine will cut the noise right back, but it can also produce some rather artificial-looking image smoothing and some loss of fine detail. To fix this, and to do some other fine-tuning, you can click the Reduce button at the top of the Noise Reduction panel.
Dfine splits noise into its two key components: Contrast noise (luminance noise) and Colour noise. Both sliders are set to 100% by default. You can safely push the Colour noise slider right up without losing any significant image detail, but most modern cameras suppress colour noise pretty well anyway.
The really tricky customer is Contrast (luminance) noise. This produces the most persistent image noise, and it’s reducing this that can take away any fine detail too and produce an artificially smooth, ‘smeary’ look.
By moving the Contrast Noise slider you can choose your own preferred compromise between noise reduction and detail rendition, but Dfine lets you go further. If you click the rather unobtrusive ‘More’ button further down the tools panel you’ll see checkboxes for Edge Preservation, JPEG Artifact Reduction and Debanding.
The Edge Preservation tool has a modest effect at preserving crisp outlines, but doesn’t help with fine, textural detail. The JPEG Artifact Reduction and Debanding options might help with particular low-quality images. None of them make a massive difference to Dfine’s abilities.
You can, however, use control points to remove or reduce the noise reduction in specific areas, preserving detail in highly detailed areas where the noise hardly shows and removing noise in large areas of even tone where it’s most obvious.
Dfine’s noise reduction is moderately effective, but it seems to have fallen behind the latest algorithms and offers only simplistic noise reduction tools. It’s worth using if you’ve got it, and can do a reasonable rescue job on noisy pictures, but it’s not the equal of MacPhun’s Noiseless, for example.