Then mathamatically it’ the same. Some ports are reduced,other parts are strengthend. What tool do you use for it?
In your dark frame example they’re subtracted.
George
Or do you mean with rgb images and layers?
This thread is about raw files.
George
This thread is about DXO being able to read the rather odd Nikon format NEFX, which is a rather Frankenstein format somewhere between RAW and, err, not RAW. That’s why it’s in Feature Requests.
NEFX files can only be opened by the RAW convertors, Capture One, Adobe Camera RAW and Nikon NX Studio, which creates NEFX files from Pixel Shift imaging.
The thread wandered into Nikon’s, and others, ability to combine multiple RAW images and make a RAW output… which kinda implies combining prior to de/mosaicing… ie not RGB.
What an NEFX file actually is, doesn’t really matter, I’d just like DXO to open it, just as Capture One and Adobe can.
I sent my latest NEFX to Capture One and they nicely sent back a screen grab showing it open in their RAW convertor software.
Just a last post.
You combine 2 identical images and averages the values so you obtain less noise… my question was 1).how do you do that and 2)how does averaging reduce noise.
George
The images are identical in as much as the camera settings are identical.
However, the actual outputs from the individual sensels are subject to the laws of physics. This results in them being more-or-less identical and that’s the point. They vary. By averaging, say 10 shots, the variable nature (noise) is mathematically reduced.
If sensors were perfect, each shot would be identical, which they most certainly are not. Some of this comes from the amplification of the tiny voltages produced leading to a spread of values. The higher the amplification, the more spread of values.
ISO 64 is a lot less noisy than ISO 10000. However, you repeat the ISO 10000 shot, say, 20 times, and average the output and you have a vastly reduced random noise image.
Your message specifies ‘two identical images’, they don’t exist, unless you rename the same image. Each will vary by a small amount… so identical doesn’t exist. Averaging smooths the messy noise.
I suggest you try this… Multiple Exposure
Ok.
You mean the in-camera multi shot. I still don’t see how averaging the layers can reduce noise. I did read Nikon is claiming that.
George
You can think of it simplified like this:
It’s statistics. Noise is some unknown signal that is added to your color information that you want to capture.
Imagine your object has a red color value of 10 and your camera sensor has a single pixel. Now your camera sensor captures the value 10, but also some noise. The noise is random, it’s like rolling a dice, one time the dice returns 1, one time 6, one time 3, etc. The noise is added to your color value, so now you captured 3 images, one returns 11 (10+1), one 16 (10+6) and one 13 (10+3). All 3 images are not correct (they should have value 10).
We do not know the correct value of 10, we only get the outputs 11, 16 and 13. We also do not know the noise value of each image, as it is random, like rolling a dice. However, we know the sensor characteristics. It’s like knowing the properties of the dice. We know a dice has 6 values, so each time you roll the dice the value will be between 1 and 6. If you roll it many times, the average value of the dice (our noise) will be 3.
Same with our sensor. If we subtract 3 from our individual images, it does not help us, as we would get 3 images with values 8 (11-3), 13 (16-3) and 10 (13-3). However, if we average all the 3 images, we get 13 ((11+16+13)/3). Now from that value 13 we can subtract our known average noise value 3, and we will receive a clean image value of 10.
Thus the more images you capture and take the average of, the more likely it is that you will receive a clean image without noise.
1 Like
When i understood you the calculation is based on a know average noise of the sensor. And probably for different iso’s.
Sounds logical. Didn’t kwo that. Thanks.
George
Yes in that example I gave that’s true. However, actually it might be even simpler as the distribution of the noise is likely centered around 0, meaning the average of the noise is 0 and you don’t need to subtract it. Taking the average of all frames is already sufficient.
If you mean the sensel values, that seems to me inmpossible unless there is no noise.
Back to the beginning again.
George
Pixel shift on the Z8 now combines with Focus shift or AE bracketing !
1 Like
This article might interest readers of this topic
1 Like
Since I shoot landscapes and cannot avoid moving water, clouds, air, etc, I can’t see me going for this. And, as for the noise reduction argument, so far, DeepPRIME seems to fit the bill and I can even shoot musicians playing on stage at up to 25,000 ISO.
There! that’s just saved me a couple of thousand Euros 
1 Like
Nice article. Thank You.
The author states…
Update:
The newer Adobe DNG converter (I’m using 16.1) DOES understand that an NEFX file is in fact a raw file, and can convert it into DNG!
Thought so! 