LR noise reduction in comparison to DPP

[bjd]

Hi,
I have a 5DIII and posted a few times lately that I am not happy with the noise visible in shots that are dark or underexposed. Underexposed especially as I took a lot of bracketed shots on my last vacation so many are at -2, -1.5 or -0.5EV. Also bought LR a few weeks ago and am busy trying to learn it.
At a special event yesterday I got a chance to show a Canon guy the noise in a few shots. He suggested the problem is in the RAW converter. So I tried out DPP and seem to have got better results. I did the following, opened the RAW in DDP and applied NR, then saved the RAW again. Then loaded the RAW into LR and developed it there.
So, is this just a fluke? Is the NR (or even the whole support for the 5DIII) better in DPP than in LR?

Also I was wondering in what order settings are applied in LR? If I increase exposure of a shot, do I amplify the noise? If I then add NR, does it apply to the unamplified noise?

Does the order make any difference at all?

Any ideas?

Cheers Brian

 

[vlad]

RAW converters certainly make a difference. Having read a few converter comparisons over the years, I do recall DPP doing better in some pixel-peeping cases than ACR/LR. It makes sense - after all, who knows the sensor better than Canon?

That said, I think that Lightroom's noise reduction is quite good. You have to play with the settings to get a feel for them, but it's great to be able to control for color and luminosity noise separately, as well as mask sharpening to prevent sharpening the noise.

The order of operations (boosting exposure vs. noise reduction) doesn't matter in Lightroom. If you reduce noise before you boost exposure, you'll probably have to reduce noise again after, since you will have revealed noise that you couldn't see before. I usually do noise reduction as one of the last steps, unless the noise is so great that it really bothers me. That way, I can make all the adjustments, see how much they've affected the noise, and then reduce noise appropriately.

Also, if I may ask, what was the subject that required exposing -2EV and then pushing it? That just seems like a recipe for noise...

Hope that helps.

 

[neuroanatomist]

Is the NR (or even the whole support for the 5DIII) better in DPP than in LR?

Quite possible. But likely it's ISO and scene dependent. Some software will handle it better than others (i.e. DPP vs. LR vs. DxO), and there are also dedicated plug-ins (Noise Ninja, Topaz Denoise). Fortunately, most offer some sort of free trial - you can test them all for yourself on the same set of images.

Also, if I may ask, what was the subject that required exposing -2EV and then pushing it? That just seems like a recipe for noise...

Perhaps a black barbeque against the side of a white house? But in that case, pushing 2 stops wouldn't be nearly enough. :

 

[Mt Spokane Photography]

One of the advantages that Lightroom has over DPP is that you can create a layer using the brush and apply NR selectively to that layer. This allows you to add more NR only to the areas that need more. Photoshop does this is a more sophisticated way, but its easy to do in LR 4.

Some of the other NR software attempts to apply NR only in the areas where its needed most, and works in some cases but not others.

You can get free downloads of most of the common NR software and try it to see how it works for you. Each has its own following.

 

[dtaylor]

In my experience Adobe's tools yield higher resolution/finer detail. Canon's DPP can be adjusted for less noise.

I use ACR for pretty much anything through ISO 800, and DPP for 1600 and 3200.

 

[bjd]

Also, if I may ask, what was the subject that required exposing -2EV and then pushing it? That just seems like a recipe for noise...

Hope that helps.

Vlad,
its mainly looking at the very underexposed ends of a bracketed sequence, and seeing that the shots are not usable for making a HDR due to the amount of noise. I find that the HDR process then amplifies the noise and the results are useless. So I need to select very carefully which shots would go into a HDR.
And therefore wondering if its correct that these shots have that much noise?
In fact I have just about given up on HDR now.

I currently have a trial version of Topaz denoise, but I dont find the results that good. Or at least
IMHO not good enough for me to buy it.

Thanks to everyone else who replied.

 

[pierceography]

Also, if I may ask, what was the subject that required exposing -2EV and then pushing it? That just seems like a recipe for noise...

Perhaps a black barbeque against the side of a white house? But in that case, pushing 2 stops wouldn't be nearly enough. :

Ahahahaha!!! I actually LOLed at that.

 

[weekendshooter]

Is the NR (or even the whole support for the 5DIII) better in DPP than in LR?

Quite possible. But likely it's ISO and scene dependent. Some software will handle it better than others (i.e. DPP vs. LR vs. DxO), and there are also dedicated plug-ins (Noise Ninja, Topaz Denoise). Fortunately, most offer some sort of free trial - you can test them all for yourself on the same set of images.

Also, if I may ask, what was the subject that required exposing -2EV and then pushing it? That just seems like a recipe for noise...

Perhaps a black barbeque against the side of a white house? But in that case, pushing 2 stops wouldn't be nearly enough. :

Serious subject matter requires serious equipment. C'mon Neuro, you of all people should know that. Now let's see your publications!

 

[vlad]

its mainly looking at the very underexposed ends of a bracketed sequence, and seeing that the shots are not usable for making a HDR due to the amount of noise. I find that the HDR process then amplifies the noise and the results are useless.

Aah I see. What software are you using for HDR? Theoretically, the HDR software should use your underexposed shots to get the highlights (i.e. sky/clouds) and your overexposed shots for the shadows (which should have little noise since they were exposed to the right originally, not pushed in post).

In practice though, it doesn't always come out as expected. I use HDR Efex, and yeah, sometimes there is noise in the "shadows", even though I can look at the overexposed shots of those areas and see that they're not noisy. I've always chalked it up to my general inexperience with HDR and possibly not having enough +EV bracketed shots...

 

[marekjoz]

(...)
I did the following, opened the RAW in DDP and applied NR, then saved the RAW again. Then loaded the RAW into LR and developed it there.
(...)

I don't think, that LR takes care about anything corrected first in DPP and the other way as well. LR looks at the source file and interpretes only settings applied in LR. LR uses separate file to store settings applied to a RAW file, while DPP writes it directly to the RAW file, but it doesn't mean, that their are anyhow interpreted by LR.
Think about settings applied in eirher LR or DPP as a "recipe". Converter (either DPP or LR) looks for this recipe each time when you open a file, takes the original file and applies those "recipe settings" to present you the result. If you are happy with the result on preview, and decide to develop the file, then the converter takes this recipe again and applies to the original RAW file to produce the result file, either jpg or tiff.
Noise settings are part of this recipe. Even if LR would read amount of noise reduction set in DPP, it applies it's own algorithm, so the result will be different.

 

[drjlo]

I really like Imagenomic Noiseware, which I use as photoshop plugin. It gives you many different ways to reduce noise, and I find the NR cleaner without killing off too much detail.

 

[bjd]

Aah I see. What software are you using for HDR? Theoretically, the HDR software should use your underexposed shots to get the highlights (i.e. sky/clouds) and your overexposed shots for the shadows (which should have little noise since they were exposed to the right originally, not pushed in post).

In practice though, it doesn't always come out as expected. I use HDR Efex, and yeah, sometimes there is noise in the "shadows", even though I can look at the overexposed shots of those areas and see that they're not noisy. I've always chalked it up to my general inexperience with HDR and possibly not having enough +EV bracketed shots...

Thats what I thought too. The noise is there, but doesn't matter as it hasn't been pushed.
I've tried with EasyHDR, Photomatix and HDRExpose. As I wasn't happy with the results, and I tried many permutations, I started to look at the source material in more detail.

 

[bjd]

(...)
I did the following, opened the RAW in DDP and applied NR, then saved the RAW again. Then loaded the RAW into LR and developed it there.
(...)

I don't think, that LR takes care about anything corrected first in DPP and the other way as well. LR looks at the source file and interpretes only settings applied in LR. LR uses separate file to store settings applied to a RAW file, while DPP writes it directly to the RAW file, but it doesn't mean, that their are anyhow interpreted by LR.
Think about settings applied in eirher LR or DPP as a "recipe". Converter (either DPP or LR) looks for this recipe each time when you open a file, takes the original file and applies those "recipe settings" to present you the result. If you are happy with the result on preview, and decide to develop the file, then the converter takes this recipe again and applies to the original RAW file to produce the result file, either jpg or tiff.
Noise settings are part of this recipe. Even if LR would read amount of noise reduction set in DPP, it applies it's own algorithm, so the result will be different.

Yeah Brain-Check at my side considering the first point, I didn't know that the DPP "recipe" would be ignored by LR, and vice-versa too probably.
Cheers Brian

 

[K-amps]

I currently have a trial version of Topaz denoise, but I dont find the results that good. Or at least
IMHO not good enough for me to buy it.

On the other hand, I tend to prefer Topaz in many situations. I find it is good for a near complete elimination of Noise in portraits and even sceneries/ city scapes. For when I want to keep more detail but can live with a bit of fine noise, I use Nik Dfine. When I am plain lazy, I use LR or when I want a slight NR applied to 200 shots at once. I never use LR for radical NR since it makes things lose too much detail and seems a bit plasticky...

Horses for courses.

 

[Mt Spokane Photography]

Yeah Brain-Check at my side considering the first point, I didn't know that the DPP "recipe" would be ignored by LR, and vice-versa too probably.
Cheers Brian

I believe that, when in DPP, you click export to Photoshop, a tif file with your DPP edits baked in, will be opened in Photoshop with the edits.
If you use the original cr2 file in Lightroom or Photoshop, you won't see the DPP edits.
Its a subtle difference that depends on the method you use to get a file to Lightroom, if you send it as a tif file it will have DPP edits baked in.

 

[kubelik]

I've found in my personal experience that Neuro is exactly right - it depends on the photo. I've done a bunch of side-by-side testing between ACR, DPP, and Imagenomic's photoshop plug-in, and which one I perfer tends to vary by the image.

so at the end of the day, I've given up on worrying about minute differences in favor of just having a consistent workflow.

 

[bycostello]

why are you underexposing? 50% of the file data is in the top 1/5th of the highlight end of the image.