Is basing noise profiles on ISO speed for a given camera, and not conssidering shutter speed at all, completely correct?
It seems to me that a night shot will never have the same noise levels as a day shot, even if you compensate using a long exposure time.
However, over some range (maybe 1/30 to 1/2000?) noise might be independent of shutter speed.
Would it be worth while to do profiles for long shutter times? Or is it really sufficient just to profile as a function of ASA?
is noise truly independent of shutter speed?
yes, noise is independent of shutter speed
For the record, after I asked the question I went out and measured 30-40 images taken under different lighting conditions, shutter speeds, etc. I use the profile viewer to look at the results. I used a spreadsheet to enter all the profile data, and then found the average results.
There was NO statistically significant difference in noise levels from shutter speeds from 1 second to 1/1000 of a second. That put to rest an old myth! That is, each profile was slightly different, probably some randomness involved , but noise didn't follow shutter speed at all. I did this many times and averaged the results, so I'm pretty confident in this conclusion.
I suppose that for multi-second exposures there might be some difference, but for any practical purposes one noise profile for all shutter speeds is sufficient - at least for my camera!
I then went further and ran some profiles with the image deliberately under exposed by 1-2 stops. You might want to do this to maximize shutter speeds - for an indoor basketball game. This basically forced Neat Image to look harder in the shadow areas. When I did this there was some slight effect of shutter speed on noise - but is was very minor, not enough for me to worry about. Could have just been random as well - when you do this this noise levels increase a lot, and the variance from shot to shot goes up at the same time.
So, for the extreme shadow areas there may be some slight dependence of noise on shutter speed, which may be what some people notice. For a properly exposed picture there is none.
There was NO statistically significant difference in noise levels from shutter speeds from 1 second to 1/1000 of a second. That put to rest an old myth! That is, each profile was slightly different, probably some randomness involved , but noise didn't follow shutter speed at all. I did this many times and averaged the results, so I'm pretty confident in this conclusion.
I suppose that for multi-second exposures there might be some difference, but for any practical purposes one noise profile for all shutter speeds is sufficient - at least for my camera!
I then went further and ran some profiles with the image deliberately under exposed by 1-2 stops. You might want to do this to maximize shutter speeds - for an indoor basketball game. This basically forced Neat Image to look harder in the shadow areas. When I did this there was some slight effect of shutter speed on noise - but is was very minor, not enough for me to worry about. Could have just been random as well - when you do this this noise levels increase a lot, and the variance from shot to shot goes up at the same time.
So, for the extreme shadow areas there may be some slight dependence of noise on shutter speed, which may be what some people notice. For a properly exposed picture there is none.
Thank you for telling about your results. Yes, it is quite interesting that the shutter speed does not significantly affect the noise levels. Hot pixels of course are much more affected but general noise that Neat Image reduces does not change a lot. This may depend on a camera too but usually the influence of shutter speed is not signifact according to our experience.
Vlad
Vlad
-
Bryan Stone
- Posts: 20
- Joined: Sat Aug 09, 2003 2:48 am
So is there no need to build a lot of profiles?
When I got NI 4.0 the other day, I redid my profiles. I did over 8 shutter speeds for each of 5 iso's 49 in all This was just for compression Fine on my A2.
It really is not all that much trouble to do these, especially with your new batch profile builder.
Since I also use Dalibor's MRWFormat, and it has multiple interpolation methods which produce different noise patterns, I had been thinking that I needed to do something about that too, at some time in the future.
Is all this overkill? Would it work just as well to have only one, two or three shutter speeds for each iso?
Best wishes
It really is not all that much trouble to do these, especially with your new batch profile builder.
Since I also use Dalibor's MRWFormat, and it has multiple interpolation methods which produce different noise patterns, I had been thinking that I needed to do something about that too, at some time in the future.
Is all this overkill? Would it work just as well to have only one, two or three shutter speeds for each iso?
Best wishes
Bryan
Bryan,
Usually one shutter speed is quite enough, especially if you additionally auto tine-tune loaded profile to currently processed image.
In principle, shutter speed may have some influence but not always, not with every camera. I saw cameras where the shutter speed did not change noise levels, and cameras where changes were evident. I believe this all depends on the way a specific camera works with the signal from its image sensor.
In any case, Neat Image takes the shutter speed into account in profile matching and selects the closest profile if you have several profiles with different shutter speeds. So, it will not hurt. But to minimize your efforts, first check how much different the noise levels are with different shutter speed. Use Profile Viewer for that. If there is no difference, then save time and efforts and build just one profile.
Hope this helps.
Vlad
Usually one shutter speed is quite enough, especially if you additionally auto tine-tune loaded profile to currently processed image.
In principle, shutter speed may have some influence but not always, not with every camera. I saw cameras where the shutter speed did not change noise levels, and cameras where changes were evident. I believe this all depends on the way a specific camera works with the signal from its image sensor.
In any case, Neat Image takes the shutter speed into account in profile matching and selects the closest profile if you have several profiles with different shutter speeds. So, it will not hurt. But to minimize your efforts, first check how much different the noise levels are with different shutter speed. Use Profile Viewer for that. If there is no difference, then save time and efforts and build just one profile.
Hope this helps.
Vlad
Julio,
That's exactly what I did with my 10D profiles (http://www.neatimage.net/forum/viewtopic.php?t=699), but using C1 as the raw convertor. I'm not sure how much extra benefit there is (haven't had time to measure), but it's easy to do this with 4.0's batch profile generator, so I figured, why not?
That's exactly what I did with my 10D profiles (http://www.neatimage.net/forum/viewtopic.php?t=699), but using C1 as the raw convertor. I'm not sure how much extra benefit there is (haven't had time to measure), but it's easy to do this with 4.0's batch profile generator, so I figured, why not?
NeatImage Pro Plus 5.0 + dual Opteron 244 + Windows XP SP2 + FreeBSD 5.2
did this as well
It's kind of like push processing film. Underexposose, and then bring the levels up in post processing. This creates a lot of noise, so I figured why not make a special profile for this. Took picture of the test image using EV -1 and -2, then opened in Photoshop and adjusted the levels back up (so greys were still grey). THEN I saved the image and used NI to profille the test image.
Now, lets say I'm taking indoor basketball pictures, I can go to ISO 400 to use a higher shutte speed, underexposose by two steps, use levels to bring everyting up, then use this special Neat Image file to bring noise back down. Not the best pictures, but they are the best pictures considering the circumstances!
Now, lets say I'm taking indoor basketball pictures, I can go to ISO 400 to use a higher shutte speed, underexposose by two steps, use levels to bring everyting up, then use this special Neat Image file to bring noise back down. Not the best pictures, but they are the best pictures considering the circumstances!
I too use MRWFormat for converting my RAW files (Minolta A1). I have found the new AutoProfile feature to be so good, that I am not using my profiles any more. This allows me to change interpolation algorithms and still get a good noise reduction without building 30 or so profiles.
First shot, I use PS CS to batch process all my images. Obviously, for this, MRFormat is set to my default settings. I still use the AutoFormat feature for this. It takes a little longer, but most of my images do not need to be tweaked any further than this.
John Fagerberg
First shot, I use PS CS to batch process all my images. Obviously, for this, MRFormat is set to my default settings. I still use the AutoFormat feature for this. It takes a little longer, but most of my images do not need to be tweaked any further than this.
John Fagerberg
It seems that temperature is also important, but no exif information for that. 
http://www.dpreview.com/news/0005/00050 ... rature.asp
http://www.dpreview.com/news/0005/00050 ... rature.asp
EXIF has some fields that could be directly used for the purposes of external noise reduction but, unfortunately, camera manufacturers seem to be too busy with other issues.
Regarding dependence of noise on temperature and other factors - these can be compensated by either auto profiling or auto fine-tuning applied before noise reduction in Neat Image.
Vlad
Regarding dependence of noise on temperature and other factors - these can be compensated by either auto profiling or auto fine-tuning applied before noise reduction in Neat Image.
Vlad