NI after RAW development

questions about practical use of Neat Image
Post Reply
Aqua2102
Posts: 5
Joined: Sun Mar 20, 2005 1:20 pm

NI after RAW development

Post by Aqua2102 »

Hi NI team,

I am a happy NI user for several years now. Let me take this opportunity to thank you for your product.

I have a few questions regarding denoise with RAW workflow. I know that this topic has already been adressed in other threads here. But while I found the info therein quite useful, I still did not fully understand how to optimally make use of NI within my RAW workflow context. Hopefully you'll forgive me in case I am asking questions you long thought answered already.

I use Adobe-Camera-Raw (ACR) for RAW development:
* The basics tab with slider based adjustments of whitebalance, exposure, brightness, contrast, saturation, etc
* The curves tab with tone adjustments using the in/out curves,
* The details tab with denoise and sharpening


I understood from previous posts that when applying NI after RAW development there are basically two ways to deal with potential changes in noise characteristics caused by different development settings:
a) Use the autoprofile option, which will try to create a dedicated profile from scratch for each new image. The potential disadvantage here seems to be that there has to be a suitable region for analysis in each image.
b) Use profile matching with auto-fine-tuning with profiles that were created with ACR settings as close as possible to those used for the real photos. The drawback here is of course, that the images the profiles are applied to may have matching camera settings (ISO, shutter speed, etc) but will usually differ in the ACR development settings.


Apart from my questions below, is there a clear recommendation which of the two approaches to use ?


1) If I use approach a):
Is there a possibility that NI does not find a suitable region for noise analysis at all ?
In that case (or if it only finds one of low quality), is there a possibility to tell NI to try profile matching as a fall back ?
Is there maybe a possibility for NI to try both approaches and use the one with the better quality ?

2) If I use approach b):
Can I expect auto-fine-tune to be able to cope with changes in noise characterists due to different development settings ?
Do tone adjustments using curves pose a bigger problem for auto-fine-tuning than the slider-based adjustments ?
Would you generally recommend to never use sharpening and denoise functions prior to NI denoising ?
Or is it safe to use them with fixed values when the same values are also used during the profile creation ?

3) And generally:
Is the quality indicator a reliable tool to indicate which approach gives better denoise results ?
Say, if e.g. I submit a bunch of pictures to NI batch processing one using approach a) and one using approach b), may I use the quality values to judge which approach works better for me ?

4) An then finally I was wondering:
Why not create profiles based directly on the RAW data and also apply denoise directly to the RAW data ?
Wouldn't that be the proper place to do the denoising anyway ?
In that sense, I would really appreciate if you guys could negotiate with the ACR team to integrate your excellent denoise tool as a plugin or something.

Anyhow, your support is much appreciated.

Thanks and BR,
Holger
NITeam
Posts: 3173
Joined: Sat Feb 01, 2003 4:43 pm
Contact:

Re: NI after RAW development

Post by NITeam »

Aqua2102 wrote:b) Use profile matching with auto-fine-tuning with profiles that were created with ACR settings as close as possible to those used for the real photos. The drawback here is of course, that the images the profiles are applied to may have matching camera settings (ISO, shutter speed, etc) but will usually differ in the ACR development settings.
And that is why Auto Fine-Tuning is suggested, to compensate for those minor changes introduced by differences in ACT settings.
Aqua2102 wrote:Apart from my questions below, is there a clear recommendation which of the two approaches to use ?
There are two ways and they both can be used depending on situation. Everything depends on your preferences, needs and, for example, willingness to prepare a high-quality reusable set of profiles for your camera.
Aqua2102 wrote:1) If I use approach a):
Is there a possibility that NI does not find a suitable region for noise analysis at all ?
In that case (or if it only finds one of low quality), is there a possibility to tell NI to try profile matching as a fall back ?
Is there maybe a possibility for NI to try both approaches and use the one with the better quality ?
Yes, there is such possibility. In the current version of NI, you cannot do that automatically, but you can manually do two runs over the batch of images: first run using Auto Profile, then the second run using Auto Match for the images that remain unprocessed after the first run. We plan to further improve that aspect in the newer versions. Regarding quality, it is not easy to assess, it can be subjective too, so that is where you have to make a decision. Which is usually not difficult for human eyes, just take a look and throw away one of the versions (if you processed each image using two different methods).
Aqua2102 wrote: 2) If I use approach b):
Can I expect auto-fine-tune to be able to cope with changes in noise characterists due to different development settings ?
Yes, that is the idea behind using AFT in that workflow.
Aqua2102 wrote: Do tone adjustments using curves pose a bigger problem for auto-fine-tuning than the slider-based adjustments ?
No.
Aqua2102 wrote: Would you generally recommend to never use sharpening and denoise functions prior to NI denoising ?
Sharpening - definitely, would not use it prior any denoising. Denoise functions - most likely too, since if you use a less sensitive (to details) denoiser than it will eliminate some fine details and then NI will not be able to detect and preserve them anymore.
Aqua2102 wrote: 2) If I use approach b):
Or is it safe to use them with fixed values when the same values are also used during the profile creation ?
It is safe, but you do not generally want to sharpen the noise before reducing it, because sharpened noise is more difficult to reduce.

Aqua2102 wrote:3) And generally:
Is the quality indicator a reliable tool to indicate which approach gives better denoise results ?
Say, if e.g. I submit a bunch of pictures to NI batch processing one using approach a) and one using approach b), may I use the quality values to judge which approach works better for me ?
Not necessarily. The quality indicator indicates the quality of the noise profile. Which is not always the same as the quality of output image. For example, a very good (high quality value displayed) profile for ISO800 will not produce very good output image if the input image is ISO1600.
The (profile) quality indicator's purpose is to help you do noise profiling right, to build a better profile. Using a better profile usually leads to more accurate noise reduction but this is not necessary nor sufficient to get a good looking output image, the quality is just one component, though quite important one.
Aqua2102 wrote:4) An then finally I was wondering:
Why not create profiles based directly on the RAW data and also apply denoise directly to the RAW data ?
Wouldn't that be the proper place to do the denoising anyway ?
It would be one of possible places. NI is in another because it is much easier to work with standard RGB data rather than with numerous proprietory RAW data formats. It is nice to be able to leave support of all those formats to ACR or other RAW converters (it is their job after all) and rely on them to do their part and then do our part. In this way, everyone does what they can do best.

Thank you for your very thoughtful questions.

Kind regards,
Vlad
Aqua2102
Posts: 5
Joined: Sun Mar 20, 2005 1:20 pm

Post by Aqua2102 »

Thank you very much for your quick reply.
Keep up the good work.

BR,
Holger
Aqua2102
Posts: 5
Joined: Sun Mar 20, 2005 1:20 pm

Post by Aqua2102 »

Dear NI team,

I did forget to ask about one thing and that is localized adjustments.

As you know ACR (and also the other RAW converters) do offer the possibility to apply adjustments for color, tone, etc not only globally to the whole picture but also locally to user defined areas (adjustment brush).

My questions here would be:
1) Does NI in general assume noise characteristics apply homogeneously to the whole image ?
2) Does the auto-fine-tune adjust for localized fluctuations or does it also only compensate global deltas ?

BR,
Holger
NITeam
Posts: 3173
Joined: Sat Feb 01, 2003 4:43 pm
Contact:

Post by NITeam »

1) yes, so it is best to not locally modify the image before noise reduction
2) only global

Vlad
l_d_allan
Posts: 35
Joined: Thu Sep 24, 2009 3:00 pm

Re: NI after RAW development

Post by l_d_allan »

NITeam wrote:
Aqua2102 wrote:
Aqua2102 wrote:4) An then finally I was wondering:
Why not create profiles based directly on the RAW data and also apply denoise directly to the RAW data ?
Wouldn't that be the proper place to do the denoising anyway ?
It would be one of possible places. NI is in another because it is much easier to work with standard RGB data rather than with numerous proprietory RAW data formats. It is nice to be able to leave support of all those formats to ACR or other RAW converters (it is their job after all) and rely on them to do their part and then do our part. In this way, everyone does what they can do best.
I can understand your reluctance to deal with a bunch of proprietary RAW formats (.CR2, .NEF, etc), but what about dealing directly with Adobe's .DNG format?

Isn't there an API to work with .DNG's ?
NITeam
Posts: 3173
Joined: Sat Feb 01, 2003 4:43 pm
Contact:

Post by NITeam »

I am not sure having RAW data in .DNG format makes it significantly easier to read. It still requires RAW->RGB conversion and that is not something Neat Image is designed for.

Vlad
mronen
Posts: 1
Joined: Mon Apr 28, 2003 8:37 am

RAW Workflow

Post by mronen »

It was suggested above to save the RAW conversion to a TIFF file and then open that file in Photoshop.

Will it be advantageous to go directly from ACR to Photoshop via 'Open Image' in ACR, without saving first to an RGB format, and if so, using 16bit?

Thanks,
Moshe
NITeam
Posts: 3173
Joined: Sat Feb 01, 2003 4:43 pm
Contact:

Post by NITeam »

Yes, that could save some time and shorten the workflow.

Vlad
Post Reply