High ISO Noise Reduction

[speedneeder]

Hello,
I'm sorry if this has already been discussed...
I downloaded a trial copy of lightroom 3 today and I can say that the high ISO noise reduction with this program seems clearly superior to anything I have been able to do with GIMP or DPP. It's much faster, the restults are clearly better, and it's easy to do. Better as in elimnates MORE noise while retaining MORE detail. This is based on some test shots at ISO 6400 on my Canon 60D and was pretty impressed with the results. Perhaps I just don't know how to use DPP or GIMP very well? I'm not sure I like the interface to import and export with LR3, are there other options I should be looking at?
(currently running a windows machine).

Thanks

[russellsnr]

Hi, If you are asking NR within LR sorry can't help but there are many 3rd party NR software with free trial versions and as they are made specific for NR I would guess they would do a very good job the one I use is Dfine from NIK (http://www.niksoftware.com/dfine/en/entry.php) The other one that has good reviews is Imagenomic Noiseware (http://imagenomic.com/).
Cheers
Russ

[Dave Humphries]

My Vote is for Neat Image

Here's how to use it;
Neat Image, a simple workflow

I agree the latest version in ACR/LR is much better than the earlier ones though.

[Bobobird]

The NR in LR or ACR is pretty good but the 3rd party stuff is better.
I used NIK Dfine and Noise Ninja before but now am almost exclusively using Topaz DeNoise.

[Colin Southern]

High ISO modes don't cause noise - they "cause" a proportional reduced dynamic range -- it's that reduced dynamic range - WHEN COMBINED WITH UNDER EXPOSURE - that causes the same noise that's captured in every (RAW) shot (even at low ISOs) to now be visible.

Or put another way - "don't under-expose and 99.9% of the time you won't need any noise reduction at ANY ISO"

[speedneeder]

Colin, I'm sure I don't understand what you are saying because it is 100% true that when I shoot at ISO 6400, there is plenty of noise at any exposure!

[speedneeder]

Russel and Dave, thanks for the suggestions.
Bobo, what made you switch to topaz?
I have of that and noise ninja before, but have never used them before, only the NR included in DPP on a very limited basis.

[Colin Southern]

Colin, I'm sure I don't understand what you are saying because it is 100% true that when I shoot at ISO 6400, there is plenty of noise at any exposure!

Hi Brian,

As a "case in point", how about sending me an offending RAW shot for me to have a look at?

PS: Brian, have a read of post 10 in this thread for more of an in-depth explanation.

[shreds]

High ISO modes don't cause noise - they "cause" a proportional reduced dynamic range -- it's that reduced dynamic range - WHEN COMBINED WITH UNDER EXPOSURE - that causes the same noise that's captured in every (RAW) shot (even at low ISOs) to now be visible.

Or put another way - "don't under-expose and 99.9% of the time you won't need any noise reduction at ANY ISO"

Sure thing Colin, but these guys who are finding the need to go to 6400 ISO are almost by definition shooting in low light and so run a high risk of under exposure, hence the attendent noise.

Unless I am using it for effect, I hate noise as much as I did grain in film days. Whilst manufacturers have made big inroads into this area in recent years, the more the camera does the more we push still further.

As you say, all these programmes effectively do is map the noise inherent in all RAW images and try to eliminate it from the chosen shots.

When I used to do my own film processing, some of these ISOs available today were just undreamt of. At least we can get shots in some situations where we would have not been able to in the past. Not necessarily wow photograhically, but nevertheless they have their uses.

[Colin Southern]

Sure thing Colin, but these guys who are finding the need to go to 6400 ISO are almost by definition shooting in low light and so run a high risk of under exposure, hence the attendent noise.

Hi Ian,

I don't disagree.

I think a big part of the problem is that many photographers are mentally associating "high ISO modes with noise" (and thus almost accept it as a given) rather than associating "high ISO modes with reduced dynamic range capability" (and thus failing to appreciate the need to push the exposure "to the right" as much as possible).

If it's a high-contrast low light scene then ... yeah ... it's going to be a problem (and one will probably have to decide on the degree of tradeoff between blown highlights and noisy shadows), but if the scene is purely reflective then they just have to learn to push push push the exposure as far as they can. I wouldn't be at all surprised if many who are having issues with noise at high ISO when shooting purely reflective scenes are under-exposing by a couple of stops or more (in terms of "wasted headroom", NOT in terms of whether or not the shot looks over-exposed / washed out on the review screen).

So long as the highlights aren't blown, in a high ISO situations they're ALWAYS going to get less noise if they over-expose as much as possible, and then reduce it in post-processing. In fact, that goes for any shot (hence the ETTR theoriests) - it's just at low ISO it's not particularly critical, but a different story altogether at high ISO. Or put another way - "at high ISO settings, sloppy exposures will be severely punished". Metering for high ISO situations can be very different than at regular ISO settings.

[PhotomanJohn]

As always, Thank you Colin. It makes complete sense but I never thought of it that way and it is too bad that the issue is not commonly explained that way. I guess another way to look at is to compare it with audio. In the old days when noise was a problem you only heard it in the quiet parts not in the loud ones. It is still there in the loud parts but you just can't hear it. I think these are both the same from the point of view of the electronics.

John

[Colin Southern]

As always, Thank you Colin. It makes complete sense but I never thought of it that way and it is too bad that the issue is not commonly explained that way. I guess another way to look at is to compare it with audio. In the old days when noise was a problem you only heard it in the quiet parts not in the loud ones. It is still there in the loud parts but you just can't hear it. I think these are both the same from the point of view of the electronics.

John

No worries John,

Yes - audio is exactly the same. It's all about signal to noise ratio - the further we can get the signal away from the noise the cleaner the result will be; normally it doesn't make much difference (at low ISOs) because of the "safety margin" (in the form of available -v- needed dynamic range), but for each stop we increase the ISO, we effectively decrease the dynamic range by the same amount (because the sensor response characteristics haven't changed -- so all that's really happening is we're under-exposing by more and more, and then amplifying the analog signal by more and more to compensate. If you start looking at some of the numbers, it's a miracle that it actually works as well as it does.

And all of that aside - even assuming that one can see the noise when pixel-peeping, it's still usually not a problem because the noise is essentially sampled out when an image is down-sized for internet display, and for a regular print (a) the dynamic range of a print is so low that the noise often doesn't show, and even if it could, our eyes can't resolve it at normal print sizes. So all in all - in my opinion anyway - "high ISO noise" is far more of a problem in theory than it is in practice. Sorry if all this is sounding somewhat "evangelical", but whenever I see people talking about noise reduction, the first thing I think of is how they're probably not doing all the right things to minimise visual noise at the time of shooting.

[Bobobird]

Bobo, what made you switch to topaz?

Main reasons are that it has standard NR presets which work well. The ability to zoom to say 200%, NR for color, blue, red, luma and rgb. Additionally has various easy to use control in sub-menus.

It allows a higher degree of direct control then the others.

[speedneeder]

Interesting, thanks Bobo.

Colin, I could post a RAW file somewhere I suppose!? I think I will test your theory about exposing to the right end of the histogram first. This does seem counter intuitive to me since the single biggest reason I use higher ISO's is to decrease exposure time - now I'm supposed to increase the decreased shutter speed!?!?! Oh the confusion. Now I will be wondering if 3200 ISO at 'normal' exposure will give better results that 6400 ISO 'exposed to the right'.
Bottom line question, will this help with my son's hockey pictures or not!? :clown face:
I already load up the right side of the camera histogram when taking hockey photos (everything is white), but maybe I need to see what happens when I let some of the whites hit the edge! (this is funny to me because the rink is called The Edge).
I also need to do some comparisons of 'down sized' .jpg's too, as I do agree with your point about high ISO noise being overly obsessed about.

Thanks

[Colin Southern]

Colin, I could post a RAW file somewhere I suppose!? I think I will test your theory about exposing to the right end of the histogram first. This does seem counter intuitive to me since the single biggest reason I use higher ISO's is to decrease exposure time - now I'm supposed to increase the decreased shutter speed!?!?! Oh the confusion. Now I will be wondering if 3200 ISO at 'normal' exposure will give better results that 6400 ISO 'exposed to the right'.
Bottom line question, will this help with my son's hockey pictures or not!? :clown face:
I already load up the right side of the camera histogram when taking hockey photos (everything is white), but maybe I need to see what happens when I let some of the whites hit the edge! (this is funny to me because the rink is called The Edge).
I also need to do some comparisons of 'down sized' .jpg's too, as I do agree with your point about high ISO noise being overly obsessed about.

Hi Brian,

At the end of the day, just about everything about photography ends up being a compromise. If one asks the question "between a slow shutterspeed or a high ISO, which will damage an image more" then if it's stopping power you need then "hands down" a slow shutter speed is going to give you more motion blur. I'd suggest just doing some formal experiments, and see what works for you; if nothing else, you might just find the justification for that super long & fast new lens that you know you want

[speedneeder]

Hi Brian,

At the end of the day, just about everything about photography ends up being a compromise. If one asks the question "between a slow shutterspeed or a high ISO, which will damage an image more" then if it's stopping power you need then "hands down" a slow shutter speed is going to give you more motion blur. I'd suggest just doing some formal experiments, and see what works for you; if nothing else, you might just find the justification for that super long & fast new lens that you know you want

LOL Colin! Of course I would love a long fast lens! My 70-200 f4 ain't bad though
Even though I use that lens at f/4 often shooting hockey, I never find myself wishing I could do it at f2.8 and actually half the time would prefer to shoot at f/5.6. This leads me to my quest for higher iso images. I find that at ISO 1600 I am happy with the photos but at ISO 3200 it is hit and miss.
You are right, it is always a compromise
I'll get back to this thread with some high iso exposure tests.

[Eric M]

I'm still impressed with DPP. I shoot RAW and leave the NR set to "standard" on my 50D. Here is a shot from last week in an underground coal mine, ISO3200. Noise reduction is automatically applied when the RAW file is viewed at 100%.

[speedneeder]

Eric,
Nice photo! Thanks for sharing.
I agree that DPP doesn't do a bad job, though it is pretty slow on my computer.
I have to apply and undo a few times to get results that I like. That is why I am impressed with LR3 - virtually instant results when making adjustments.

[speedneeder]

In an effort to beat this dead horse...

I took a couple photos tonight, thought some grass might be good to use, lots of lines, shaded areas, areas of same color, etc.
In all of these examples, there is no editing at all other than NR in some as explained (ie. no contrast, saturation, or sharpness adjustments, though LR3 seems to have done something with color saturation secretly?) Everything shot at ISO 6400 on my Canon 60D with my 70-200 f/4 L IS.

Photo 1 - original RAW file dumped to jpeg and resized to 1024 wide before uploadig to flickr (which does it's own compression I believe).


Photo 2 - some DPP NR, dumped to jpeg, resized to 1024, uploaded to flickr.


Photo 3 - LR3 NR, dumped to jpeg, resized to 1024, uploaded to flickr.


Photo 4 - similar content, +1 EV when taken, reduced -1 EV in DPP, dumped to jpeg, resized to 1024, uploaded to flickr.


Photo 5 - similar content, +2 EV (histogram loaded to the right edge), reduced -2 EV in DPP, dumped, resized, flickr'ed.


Thoughts?

[speedneeder]

Photo 6 - raw converted to jpeg, resized, flickr'ed


Photo 7 - with DPP NR


Photo 8 - with LR3 NR