PS frame averaging considerations

[chrismuc]

Some considerations regarding 'frame averaging'. This Phase One feature seams popular and well working.

I want to recall two (photoshop) methods introduced by forum fellows possibly eight to ten years ago which help to reduce noise in dark areas of pictures by digital cameras.
(this topic does not discuss the other frame averaging benefit of simulating a long exposure time)



1.
The first method seams rather similar to the Phase One idea (please correct me if I am wrong), just using PS.

- take 2/ 4/ 8 or more pictures of correct exposure
- after applying the same raw conversion, load them into layers of one PS file
- reduce the opacity of each upper second layer to 50%
- always merge that pair of layers to one layer
- proceed that process until only one layer left

Enclosed a small part of a GFX100 test file.
To enhance the effect, I underexposed the picture by two stops, then in ACR added two stops exposure brightness and I lifted the shadows by 50%.
This results in somehow grainy dark (lifted) areas.

Following are
- original picture
- average of two frames
- average of four frames
- average of eight frames

The more frames averaged, the less noisy are the dark areas.
Disadvantages:
- rather long process due to the today's large high resolution files
- leaves and other moving subjects will become blurry to a certain extend

[chrismuc]

2.
This method only requires two pictures to take.
One normally exposed and one for example three steps overexposed.
The idea is to use the normal exposed image for the medium/bright areas (which are already essentially noise free so no frame averaging or other process required) and the overexposed for the dark areas.

I used for my test one of the pictures from the first test (two stops underesposed) plus one picture one stop overexposed. Means three stops difference in exposure between the two pictures.

The first one, like above, I added in ACR two stops exposure brightness and I lifted the shadows by 50%.
The second one, I reduced the exposure in ACR by one stop and also lifted the shadows by 50% to achieve an image of generally same exposure and appearance.

First image in lower layer of PS file, second (originally overexposed) in upper layer.
Select
blend mode: normal
opacity: 100%
blending options: blend if grey
change from 0 ... 255 to for example 0 ... 100

That means that only the dark areas of the upper layer (which are virtually noise free due to the three stops overexposed and then reduced brightness) are used from the upper layer and they are blended into the lower orginal file.

The result is similar or even a bit better (less noise) than the 8x frame averaged image from the test above.

Therefore I consider this method pretty effective and quick.

I used it long ago with 5D and 5D II files, since the advanved Sony sensor technology took over in most cameras, I forgot about it. But it's still usefull, so I think I will give it a try more often from now on (in tripod pictures without moving subjects).

[SrMi]

Some considerations regarding 'frame averaging'. This Phase One feature seams popular and well working.

I want to recall two (photoshop) methods introduced by forum fellows possibly eight to ten years ago which help to reduce noise in dark areas of pictures by digital cameras.
(this topic does not discuss the other frame averaging benefit of simulating a long exposure time)



1.
The first method seams rather similar to the Phase One idea (please correct me if I am wrong), just using PS.

- take 2/ 4/ 8 or more pictures of correct exposure
- after applying the same raw conversion, load them into layers of one PS file
- reduce the opacity of each upper second layer to 50%
- always merge that pair of layers to one layer
- proceed that process until only one layer left

Enclosed a small part of a GFX100 test file.
To enhance the effect, I underexposed the picture by two stops, then in ACR added two stops exposure brightness and I lifted the shadows by 50%.
This results in somehow grainy dark (lifted) areas.

Following are
- original picture
- average of two frames
- average of four frames
- average of eight frames

The more frames averaged, the less noisy are the dark areas.
Disadvantages:
- rather long process due to the today's large high resolution files
- leaves and other moving subjects will become blurry to a certain extend

A simpler approach than handling opacity manually is to turn all layers into a Smart Object and apply stack mode "Mean". Often you can take the images handheld and use Photoshop's "Auto-Align Layers" to compensate for any camera movement.