Night Photography Reducing Noise

[xpatUSA]

The GIMP manual gives actual formulae for blending modes, does Adobe?

Here's the GIMP additive description, FWIW:

Addition mode is very simple. The pixel values of the upper and lower layers are added to each other. The resulting image is usually lighter. The equation can result in color values greater than 255, so some of the light colors may be set to the maximum value of 255.
The equation is:

E = min(M+I, 255)

The mode is commutative; the order of the two layers doesn't matter.

From https://docs.gimp.org/en/gimp-concepts-layer-modes.html

Where M=mask, I=image

Don't know if it can be applied to more than two layers ...

[Stagecoach]

Here's the GIMP additive description, FWIW:

It's apparent that the PS 'Summation' mode is one of the few that Adobe does not give details on.

I have found a couple of descriptions from non Adobe sources;

a) Summation = The sum channel values for all non transparent pixels.

b) Summation : This option calculates the pixel values from all images in the stack. This increases the resolution in the more faint areas of the images, but the brighter areas just become white.

I suppose I could do a few identical pixel value measurements between one image and the result of a stack of say 2 or 3 to see if this confirms anything.

[xpatUSA]

I have found a couple of descriptions from non Adobe sources;

<>

b) Summation : This option calculates the pixel values from all images in the stack. This increases the resolution in the more faint areas of the images, but the brighter areas just become white.

"faint areas" and "resolution" seem a bit vague but there might be a clue there if taken literally?

I suppose I could do a few identical pixel value measurements between one image and the result of a stack of say 2 or 3*** to see if this confirms anything.

... and I imagine it would help if that stack were of exact duplicates so you could hone in one pixel's xy coordinates.

How was the egg, Granny?

[xpatUSA]

... and I imagine it would help if that stack were of exact duplicates so you could hone in one pixel's xy coordinates.

Out of interest, I took a dark image into the GIMP. On a leaf at x,y = 799,403, RGB was 070, 134, 053.

I duplicated it in 3 layers, and set those layers' mode to additive. I expected that an export of all layers visible would get me 4x the brightness of the base layer but it was not to be.

I exported 4 images: base vis. 2 vis. 3 vis and 4 vis. ...
The green values ('V' in HSV for the leaf) were 134, 181, 217, 250 respectively. Definitely non-linear - no idea why yet.

E = min(M+I, 255)
From https://docs.gimp.org/en/gimp-concepts-layer-modes.html

Can't reconcile that with the numbers I got, grump.

[xpatUSA]

It's apparent that the PS 'Summation' mode is one of the few that Adobe does not give details on.

I have found a couple of descriptions from non Adobe sources;

a) Summation = The sum channel values for all non transparent pixels.

b) Summation : This option calculates the pixel values from all images in the stack. This increases the resolution in the more faint areas of the images, but the brighter areas just become white.

Just read this in my earlier link:

"Note:

Prior to blending, images have gamma correction removed and are converted from sRGB to linear. After blending, they are converted back from linear to sRGB."

So "addition" in the GIMP involves exponentiation!

Let's say V=100 and also 100 in the layer above. Let's take sRGB gamma as 2.2, even though it ain't. Then:

(100/255)^2.2 + (100/255)^2.2 = 0.25506. Then 0.25506^(1/2.2) = 0.5374 x 255 = a value of 137.

So, in the GIMP the values are not simply added up and my results in #44 make sense.

Adobe? Anybody?

[Stagecoach]

Just read this in my earlier link:

"Note:

Prior to blending, images have gamma correction removed and are converted from sRGB to linear. After blending, they are converted back from linear to sRGB."

So "addition" in the GIMP involves exponentiation!

Let's say V=100 and also 100 in the layer above. Let's take sRGB gamma as 2.2, even though it ain't. Then:

(100/255)^2.2 + (100/255)^2.2 = 0.25506. Then 0.25506^(1/2.2) = 0.5374 x 255 = a value of 137.

So, in the GIMP the values are not simply added up and my results in #44 make sense.

Adobe? Anybody?

It's also not linear in PS Ted.

From a similar test using identical images;

1 image 39,38,38
2 stacked 156, 150, 149
3 stacked 181, 179, 179
4 stacked 201, 204, 204

The jump between 1 and two images is significant.

[DanK]

So probably gamma encoding, but more than that. One issue in summing would be what to do when you exceed 255. Censoring would be a simple option, but it would distort the image. So perhaps they are applying some form of concave-down curve that asymptotes at 255.

My takeaway so far is that I am not going to use this.

It's also not linear in PS Ted.

From a similar test using identical images;

1 image 39,38,38
2 stacked 156, 150, 149
3 stacked 181, 179, 179
4 stacked 201, 204, 204

The jump between 1 and two images is significant.

[xpatUSA]

It's also not linear in PS Ted.

From a similar test using identical images;

1 image 39,38,38
2 stacked 156, 150, 149
3 stacked 181, 179, 179
4 stacked 201, 204, 204

The jump between 1 and two images is significant.

Significant enough to diverge considerably from the GIMP. To eliminate a variable or two, I made a new image, grayscale, 41/255. Duped it X3.

They went, starting at base:

41, 59, 73, 84 according to two different apps' color-pickers.

Additionally, I did the theoretical GIMP calc using 1/2.2 gamma and got:

41, 56, 68, 77 ... close enough for Government Work.

As you say, the Elephant in the Room is doing something most odd.

[Stagecoach]

So probably gamma encoding, but more than that. One issue in summing would be what to do when you exceed 255. Censoring would be a simple option, but it would distort the image. So perhaps they are applying some form of concave-down curve that asymptotes at 255.

My takeaway so far is that I am not going to use this.

I suppose it's usefulness depends upon what 'constraints' you have with respect to total exposure time of the scene.

As an example, to get the most noise free picture at night of a 'scene' in camera with a single exposure which is going to be at base ISO, exposure time may be 120s. It may be that 120s gives you too much movement on say vegetation, water, clouds, car light trails or a ship swinging on anchor.

If our preferred total exposure time is say 15s the possibilities of achieving this can be done in two ways, a single shot with higher ISO, or, multi shots that are stacked at a lower ISO. It's these two methods I'm hoping to compare when done in camera (multi exposure) or in PS stack (summation) mode.

[Stagecoach]

Significant enough to diverge considerably from the GIMP. To eliminate a variable or two, I made a new image, grayscale, 41/255. Duped it X3.

They went, starting at base:

41, 59, 73, 84 according to two different apps' color-pickers.

Additionally, I did the theoretical GIMP calc using 1/2.2 gamma and got:

41, 56, 68, 77 ... close enough for Government Work.

As you say, the Elephant in the Room is doing something most odd.

And it's getting stranger delving into PS summation mode. It now looks like I can place an almost black frame at the start of the stack set reducing that 'jump' that then enables more images to be added before reaching the required final output brightness level.

I'm going to check this again to ensure I have not made a mistake.