Workflow for (Canon) cameras with small pixels

[Glenn Pure]

I’m posting this I’ve had a few queries about my processing approach, specifically why I’m applying noise reduction to relatively low ISO images and why I use Digital Lens Optimiser (DLO). DLO is one of the tools in Canon’s RAW processing software Digital Photo Professional (DPP). It increases sharpness but at the expense of adding some noise. As this is about workflow, I’m posting it here and would value any comments.

Note that my workflow is aimed at the particular problems of cameras with small pixels – this means very high pixel count full frames like the Canon 5Ds cameras or crop sensor cameras with relatively high pixel counts like my 24 megapixel Canon 80D. At the pixel level, these are noisier and much more prone to show up sharpness and other deficiencies in the optics.

So here’s what I do. I convert my RAW files in Canon DPP where I apply DLO, invariably at the default setting of 50. DLO applies deconvolution to the RAW image. In short this means it uses proprietary data from Canon on the lens, focal length (if a zoom), focus distance, aperture and camera sensor to correct precisely for the optical artefacts they introduce. Access to DLO isn’t the only reason I use DPP for RAW processing. There is a body of evidence that DPP produces cleaner and significantly less noisy images than tools like Adobe Camera Raw. DPP and DLO enable me to get the sharpest, lowest noise images from my RAW files. But as noted, the sharpness gain from DLO is offset by some additional noise.

After generating a 16 bit TIFF from DPP, I clean up the noise in the image with Neat Image. This software works as a plugin to the Adobe products like Photoshop and Photoshop Elements. For those unfamiliar with Neat Image, it essentially provides the best of both worlds by reducing noise in a way that detail is hardly affected. There are equivalent tools like this including the free Dfine tool provided as part of the Google Nik suite but that another story. These NR tools are much more refined than generic NR tools found in common image processing software where application, particular for luminance noise usually is at considerable cost to image detail.

It’s probably more useful to jump to some examples at this point. The image below is a 1:1 crop processed through DPP and Photoshop Elements with Neat Image. The panel on the left shows no DLO and no NR in Photoshop Elements. The mid panel shows DLO but no Neat Image NR. The right panel shows DLO combined with Neat Image NR. In this case, the NR has been applied globally and at very modest levels designed to treat noise without significantly affecting detail.


For posting heavily downsided images, as required on BPN, the effects seen on a 1:1 crop are not going to be as apparent so I’ve done a similar set of three panels (in the same order as above: No DLO and no NR, DLO only, DLO and NR). In this case, I’ve applied very modest NR to the bird and much stronger NR to the background.


In the above case, the benefits of applying DLO and Neat Image NR are harder to see but I still think the image on the right is noticeably better than the left panel. Note how good detail is still preserved in the in-focus leaf just above the bird’s head even though it has had much stronger NR applied. The other benefit of applying NR in the way I have here is that it considerably reduces the amount of compression required for jpg images posted to BPN. Even modest noise especially in the background generates much larger jpg files and results in stronger compression being needed to keep files under 400KB file limit on BPN.