I am still working on that thing and looking at many different things.
If you use high ISO or long exposure you will create noise so....
Well, turns out that if you know that you will create noise (or that you camera will)...
PLAN FOR PP!!!!
Process:
Take two or more pictures, same exposure, same focus.
File script Load files onto stack
Edit - auto align
Edit - auto blend
Flatten the images
You are done.
Simple as that. The more pictures you have the better the result.
Crazy isn't it???
And, oh, by the way, if you do a real dof stacking at the same time??? This works too. I have not tried with some maniacal HDR but I would not doubt that it will work in that case too but then why use high ISO?
Rongnongno wrote:
I am still working on that thing and looking at many different things.
If you use high ISO or long exposure you will create noise so....
Well, turns out that if you know that you will create noise (or that you camera will)...
PLAN FOR PP!!!!
Process:
Take two or more pictures, same exposure, same focus.
File script Load files onto stack
Edit - auto align
Edit - auto blend
Flatten the images
You are done.
Simple as that. The more pictures you have the better the result.
Crazy isn't it???
And, oh, by the way, if you do a real dof stacking at the same time??? This works too. I have not tried with some maniacal HDR but I would not doubt that it will work in that case too but then why use high ISO?
I am still working on that thing and looking at ma... (
show quote)
Basically, noise cancels itself out through stacking. I have used this method of stacking in the past when doing scans with a cheap film scanner, to diminish the digital noise the scanner produced. Works surprisingly well, actually. Noise reduction plugins tend to destroy image detail when used aggressively, but stacking does not. For every stack layer, you must adjust the transparency in order to retain the original brightness. 1st layer, 100%. 2nd layer, 50%. 3rd layer, 25%. 4th layer, 12.5% - halving the amount of the previous layer. I usually did between 4 and 6 layers, depending on the image.
Rongnongno wrote:
... but then why use high ISO?
That's precisely why you should not waste your time exploring noise that you create artificially. You are creating noise in a manner that could easily be avoided. It is much better to avoid noise in the first place by overwhelming it with a strong signal than to try and remove it later in post processing. Also, stacking only works for stationary subjects, like when scanning film.
Stacking can average out totally random noise, but not all noise is random. For example, noise from long exposure is likely to remain although much of it can be easily removed in the camera.
Only 1/4 of the information in the raw file is for pixels under red filters of the Bayer array. The reason you saw red noise more easily is simply because it is easier to distinguish visually from the blue and green noise, but half of the luminance noise is from pixels with green filters and the other 1/4 is from the pixels with blue filters. But by the time you see them in a TIFF, JPEG or PNG, all of those pixels have been combined during raw conversion and demosaicing (see
http://www.cambridgeincolour.com/tutorials/camera-sensors.htm ) into RGB pixels, so some of the luminance noise has already been averaged out.
Knowledge is power. The more you know about the various types of noise, the less inclined you will be to try and "fix" it by brute force with convoluted manual post-processing steps.
You are going to have to concede that the people who create noise reduction software (stand a lone or plug in) know a lot more about noise than you ever will. Until you reach that conclusion you are just wasting your time.
Rongnongno wrote:
... Take two or more pictures, same exposure, same focus.
File script Load files onto stack
Edit - auto align
Edit - auto blend
Flatten the images
You are done.
Simple as that. The more pictures you have the better the result.
Crazy isn't it??? ...
Crazy, yes!
If you have time for two images, why not simply one image at half the ISO and half the shutter speed (twice as long)? Or instead of four images, one image at 1/4 the shutter speed (4 times as long) and 1/4 the ISO?
It takes less time and you don't need any post-processing to stack the images.
To all readers:
Please disregard all posts from selmlie.
(And the reactions to ths post).
rook2c4 wrote:
Basically, noise cancels itself out through stacking. I have used this method of stacking in the past when doing scans with a cheap film scanner, to diminish the digital noise the scanner produced. Works surprisingly well, actually. Noise reduction plugins tend to destroy image detail when used aggressively, but stacking does not. For every stack layer, you must adjust the transparency in order to retain the original brightness. 1st layer, 100%. 2nd layer, 50%. 3rd layer, 25%. 4th layer, 12.5% - halving the amount of the previous layer. I usually did between 4 and 6 layers, depending on the image.
Basically, noise cancels itself out through stacki... (
show quote)
I gave the automated method yours, I was not aware of, I will have to try it. :lol:
Rongnongno wrote:
To all readers:
Please disregard all posts from selmlie.
(And the reactions to ths post).
If your objective is to play with post processing without learning or accomplishing anything useful, feel free to continue.
Most members are already ignoring your "noise" posts. There are a lot of people with common sense out there.
I was in NYC and shot some night shots at ISO 1600 and they are extremely noisy.
Topaz de noise only made the overall image soft.
For some reason I did bracketing when I took those shots so I just tried the stacking procedure and now they look much better and useable.
Rongnongno, I did try the action but was not sure how to do the adjustments.
Overall the image looked better but still had unacceptable noise.
Great tip on stacking, I have common sense and follow any tips to reduce noise or improve sharpening.
Thanks.
lorenww wrote:
Rongnongno, I did try the action but was not sure how to do the adjustments.
Ok, I will try to write a PDF on how to adjust the action. Can you PM me the image?
Keep up the good work on your quest to understand and refine methods to defeat digital noise in images, Rongnongno! As you, I hate digital noise and find it very ugly, particularly of the colored type with its hideous blotches of magenta, green, etc. speckled all over otherwise beautiful images.
i think stacking images to reduce noise reduces it by 1/sqrt(n) not 1/n. so 4 images would half the noise (3 dB, 1 stop), 16 for 1/4 reduction (6 dB, 2 stops), etc
Rongnongno wrote:
I gave the automated method yours, I was not aware of, I will have to try it. :lol:
This technique (image averaging) is described in the article
http://www.cambridgeincolour.com/tutorials/image-averaging-noise.htmAveraging these images reduces the most common noise sources. What's left is a trade between object motion and noise.
Rongnongno,"stack to reduce noise"... yep, sorta' sounds crazy, but it is not.
The Panasonic ZS-20 (my glued to the hip 24/7 camera) extended to 40x and in low level light"hand held" has a popup that says multi shots will be taken...Reviewer said digital 40x is excellent because it blends.
Noise is random "white noise" of electrons (radio tube analogy) boiling off. White from exposure 1, mixed with black from exposrue 2 and the net of zero. Zero being clean.
There is a similar trick using multi layers.
http://www.graphics.com/article-old/reducing-image-noise-three-ways-photoshop
selmslie wrote:
That's precisely why you should not waste your time exploring noise that you create artificially. You are creating noise in a manner that could easily be avoided. It is much better to avoid noise in the first place by overwhelming it with a strong signal than to try and remove it later in post processing. Also, stacking only works for stationary subjects, like when scanning film.
Stacking can average out totally random noise, but not all noise is random. For example, noise from long exposure is likely to remain although much of it can be easily removed in the camera.
Only 1/4 of the information in the raw file is for pixels under red filters of the Bayer array. The reason you saw red noise more easily is simply because it is easier to distinguish visually from the blue and green noise, but half of the luminance noise is from pixels with green filters and the other 1/4 is from the pixels with blue filters. But by the time you see them in a TIFF, JPEG or PNG, all of those pixels have been combined during raw conversion and demosaicing (see
http://www.cambridgeincolour.com/tutorials/camera-sensors.htm ) into RGB pixels, so some of the luminance noise has already been averaged out.
Knowledge is power. The more you know about the various types of noise, the less inclined you will be to try and "fix" it by brute force with convoluted manual post-processing steps.
You are going to have to concede that the people who create noise reduction software (stand a lone or plug in) know a lot more about noise than you ever will. Until you reach that conclusion you are just wasting your time.
That's precisely why you should not waste your tim... (
show quote)
Knowledge IS power. So it is confusing that you would claim that the more you know about a problem, the less inclined you will be to fix it. It is the exact opposite; the more you know, the more you CAN DO to fix a problem. And noise is definitely a problem with low light photography.
Many cameras will do long exposure noise reduction, but the exposures must be very long (typically >10s). This technique takes an equally long dark exposure (shutter closed) and subtracts it from the normal exposure. This mitigates the effects of some systematic noise sources (hot pixels, charge leakage) but not others (sensor pattern noise), and not random noise sources (photon, thermal, read, atmospheric).
Multiple images are a method to mitigate the effects of random noise sources, as well as non-linear systematic sources such as charge leakage.
Both techniques provide an improved final image, which can then be processed by standard noise reduction software to further enhance the image. But noise reduction on a single image is always a trade between reducing noise artifacts and sacrificing image quality. Multiple image noise reduction techniques have more information to use, and this lessens the impact on image quality.
If you want to reply, then
register here. Registration is free and your account is created instantly, so you can post right away.