"Grain Stacking" - novel idea, or waste of time?

[rodeo_joe1]

With lockdown giving me time to sit and think (always dangerous), I thought about applying oversampling noise-reduction techniques to film grain.

 

As a proof-of-concept, I shot a series of identical frames of a still life. The frames, or part thereof, were then 'scanned' with a 24 Mpx digital camera to an equivalent resolution of about 7200 ppi.

 

Here's the full frame, so that you can judge how small the cropped sample area is.

This first crop is from a single frame showing the native grain - film is T-max 100 rated 80 EI and developed in HC-110.

The grain is already pretty fine.

 

Now a combination of two identical, but separate frames.

To my eye there's a subtle improvement in the fineness of grain; about the root 2 improvement that theory predicts. Combining 4 frames would theoretically halve the grain, but what effect that would have visually I've yet to see.

 

The scanned frames were combined using PhotoShop layers, carefully aligned and using the 'Darken' blend mode. I combined them as negative images before inverting them to a positive and adjusting the tone curve. Maybe combining them after inversion would give a different - worse? - better? - result. Who knows.

 

It'll obviously only work with subjects that stand still long enough to shoot multiple frames that can be registered accurately.

 

So. Daft idea or worth pursuing? You decide.

 

I don't know how the below image got inserted, and seemingly I can't edit it out. Sorry!

 

Edited April 20, 2020 by rodeo_joe|1

[Henricvs]

I can see there is a difference. I would be interested to see how it improves with layers.

[jim_metcalfe]

Years ago, astrophotographers called this technique "integration printing". Since they were using massively pushed B/W films that yielded very coarse-grained negs, the idea was to print multiple negs of the same subject onto the same sheet of paper. The grain pattern in each neg was different, but the image was not. Needless to say, this required serious darkroom skills. I recall reading about this in either Astronomy or Sky and Telescope magazine in the seventies.

[rodeo_joe1]

I only took the sequence of identical shots to check a focus discrepancy that I noticed between my F3 and D800. Using the same lens - a 50mm f/1.4 Nikkor S.C. - the F3 best viewfinder focus showed a different distance on the lens scale than on the D800 using LiveView to focus. From the same tripod head position.

 

So I had a bunch of identical film frames with tiny, tiny, focus increments between them. My first thought was 'focus stacking', then 'hang on, that should reduce the grain (noise) as well.' And so the idea was born.

 

I'll try to find 4 frames scanned orthogonally enough to stack accurately. Any relative rotation is an absolute pig to correct. I might have to re-'scan' the frames a bit more carefully.

 

It's an interesting effect, but whether it has any useful practical application is another matter.

 

(FWIW, I haven't yet got to the bottom of the focus discrepancy. The film resolution doesn't seem good enough to see a few millimeters of focus change.)

[Dave Luttmann]

Years ago, astrophotographers called this technique "integration printing". Since they were using massively pushed B/W films that yielded very coarse-grained negs, the idea was to print multiple negs of the same subject onto the same sheet of paper. The grain pattern in each neg was different, but the image was not. Needless to say, this required serious darkroom skills. I recall reading about this in either Astronomy or Sky and Telescope magazine in the seventies.

I did it a lot with hyper sensitized Kodak PPF400 in 35mm format. Two stacked Negs reduced grain somewhat. Both hypered film and Cold camera film.

[glen_h]

It seems to me that there are many spatial filtering systems that could apply to hide grain.

 

Reminds me, though, of deconvolution, which sometimes can do what you might think could not be done.

 

In the early Hubble telescope days, with the defective mirror, deconvolution could produce sharp images from the blurry ones.

 

That works if you know very exactly the point spread function (which they did) and have enough S/N ratio.

(So they chose bright objects to image.) That might not apply to grain, though.

[rodeo_joe1]

I did it a lot with hyper sensitized Kodak PPF400 in 35mm format. Two stacked Negs reduced grain somewhat. Both hypered film and Cold camera film.

Years ago, astrophotographers called this technique "integration printing". Since they were using massively pushed B/W films that yielded very coarse-grained negs, the idea was to print multiple negs of the same subject onto the same sheet of paper.

Yes, I was pretty sure I'd read of it being used optically with grainy film or plates. I'm not claiming I invented the technique, but I don't think I've heard of it being done with scans from film.

 

I was skeptical that it would work until I tried it, due to interaction between the film grain and the digitising pixels. But I suppose that as long as the character of the noise is consistent and random between samples the noise will (partially) cancel, whether it's caused by grain, pixel aliasing or any other cause.

 

I'm working on putting more scans and layers together. I re-digitised a strip of the negatives, making sure that the film didn't skew in the carrier this time. Hopefully I'll find 4 or more frames that align properly.

 

And maybe T-max 100 wasn't the best choice of film to try the technique with!

Edited April 21, 2020 by rodeo_joe|1