Adventured in noise reduction with Affiity Photo and the Z6

[billtils]

This came about because the President of my local camera club asked me to give a talk on Affinity Photo's more advanced features. Noise reduction was one of those that I was keen to explore anyway, particularly with a stack generated from a hand-held, low light, burst in the Z6 to see if it could be used in my regular Winter and Spring shoots of a local choir in the low night-time light of Dunblane Catherdral (and taking advantage of the Z6's silent shooting mode). A preliminary image alignment step is not required as it is incorporated in the stacking software.

The set up was a vase of artificial flowers, hand held, pre-sunrise natural light coming through the window. 1/80s, f/5, and auto ISO at 12,800 (the arbitrary limit I had selected at set up of the new camera and which turned out to be about -5EV for correct exposure) and a 5fps burst of 10 shots. The "crop" below this is approximately 50% of the full frame.

1: The SOOC image






2: A crop of the flowers after exposure adjustment and before noise reduction






3: A crop of the flowers after noise reduction by stacking






4: A crop of the flowers after noise reduction using Affinity's noise reduction tool

[DanK]

As someone who doesn't use that software and has no idea what stacking algorithm the ctacking uses, I can talk only in vague generalities, but this is what I would expect. Most of the noise in the image is random, so it will average out (I'm using "average" loosely, as I don't know the algorithm) over multiple exposures, leaving the signal--the noise-free signal. In the area of the flowers, you are averaging (again, loosely speaking) the same values, so nothing changes.

In contrast, noise reduction software has to guess at what the final pixel values should be, using either an algorithm, patterns derived from machine learning, or both. Even though AI can add a great deal of power to this process, it remains a guess, just a better informed one.

What was news to me, as soimeone who uses neither that camera or that software, is that the noise generated at -5 EV is far beyond what the software can handle.

[billtils]

Dan, yes that's how it works.

I'm not sure what you mean with "the noise generated at -5EV is far beyond what the software can handle". If the software is the native AP noise reduction then I thought it did a pretty good job for ISO 12800 plus a 5 stop underexposure - not that that matters as I'm pleased with the stacking outcome and look forward to using it at this year's choral concerts.

[DanK]

Hi Bill,

I'm saying that while I would be happy with the result from stacking, I wouldn't be with the result from AP. It looks pretty much the same as the uncorrected crop. Are you sure you posted the right version?

Dan

[billtils]

Thanks Dan.

No I'm not happy with it either but to be fair I've used that program successfully for noise reduction quite a lot with more success than here. To get a better benchmark for the effectivenss of the stacking option I downloaded a trial copy of TopazDeNoise. Here is a screen grab showing the before and after with it, and a repost of the AP Stacked version

Topaz





AP Stack






Not much to choose between them, the Topaz one possibly has the edge but the AP one is free