Noise reduction; where does it fit in the workflow?

[Lani]

I have been searching for a post I think I remember seeing here that answered this question, but can't find it.
I have been having a play with neat image (trial version) and it seems to have a nice effect on portraits in particular, but I am not sure when I should be doing it. I know the general consensus is to sharpen last, but what about noise reduction, when should I be doing that?
I am using ACR, and CS3.

I have tried the CS3 noise reduction filter but am not sure if that is good enough compared to using a dedicated program.
The other one I am considering is the Dfine software, as part of a Nik software package.
Also , which version of Neat Image do you use, there is a home edition, and pro, some with the plug-in version, which I assume would be more workflow friendly?

sorry for going off on a tangent from the original question. :D

[Miaow]

If i do any I tend to do it after I've sharpened if I'm doing with with PhotoImpact mainly since I've found that sharpening can sometimes create more obvious noise.

I have Neat Image but haven't really played with it much yet

[mickeymoo21]

Hi Lani, for me nosie reduction is the final step in my workflow (well apart from resizeing for the web that is). I use neat image (trial version) as a stand alone program, I have found it to give exceptional results in most conditons. Also I am always working in tiff format and as you may know the trial version doesn't allow you to save in tiff only jpeg so therefore for me its better to work all the way though with a 16bit tiff untill the last step where it is then converted to an 8bit jpeg, of course if you were going to get serious you would want the full version so you could complete your workflow in tiff format and as a bonus your exif data would remain intact.

Michael.

[davesmith]

This is an excellent question, one I've never been able to find a good answer to either.

All the dedicated noise reduction software/plugins are relatively good at what they do. And I understand everyone will have their own methods and preferences for what seems to be the best method.

But at what point should it be used for maxmium effect, which I think is Lani's original point. It's widely accepted that sharpening is done last, but sometimes I do two sharpening routines, which doesn't go with that "flow". And I think if you apply noise reduction as a final step after sharpening it loses sharpness. From my experience it can "smudge" what you've just sharpened.

I often wonder if noise reduction should be done before anything (eg basic levels, curves etc), done mid-flow (after basic levels/curves adjustments, but before say, some saturation boost), or at the end (before final sharpening).

[ving]

PSP12 has a really good noise reduction which i use as the last step in my processing as it has an attached sharpener too. some times i do some spot sharpening after tho.

[mercho]

I use noise ninja which is an awesome noise reduction program, but i usually find it softens the image after the noise reduction process...

So for me it really depends on the image when it comes to noise reduction, some times ill do it last, some times ill sharpen first....

very very good question though :) ill be interested to see some more of the answers...

[Lani]

T

I often wonder if noise reduction should be done before anything (eg basic levels, curves etc), done mid-flow (after basic levels/curves adjustments, but before say, some saturation boost), or at the end (before final sharpening).


Yes Dave, that was what I was thinking too, and I guess it would really depend on the image, and the effect you were after. A bit of softness with portraits isn't really a bad thing, but for landscape and architectural type stuff, I agree it would defeat the purpose of sharpening. Ahhh, another thing to try and work out. :D

Thanks for the responses everyone. :)

[Lani]

Hi Lani, for me nosie reduction is the final step in my workflow (well apart from resizeing for the web that is). I use neat image (trial version) as a stand alone program, I have found it to give exceptional results in most conditons. Also I am always working in tiff format and as you may know the trial version doesn't allow you to save in tiff only jpeg so therefore for me its better to work all the way though with a 16bit tiff untill the last step where it is then converted to an 8bit jpeg, of course if you were going to get serious you would want the full version so you could complete your workflow in tiff format and as a bonus your exif data would remain intact.

Michael.

Hi Michael,

just a bit OT, but what are the advantages of saving in Tiff as opposed to PSD, which is what I have been doing, up until I convert to jpeg before printing/web publishing.? Is there an advantage or is just personal preference?

[ving]

I often wonder if noise reduction should be done before anything (eg basic levels, curves etc), done mid-flow (after basic levels/curves adjustments, but before say, some saturation boost), or at the end (before final sharpening).
i vote dave most quotable person in this thread :)

noise reduction last... why?
adjusting levels, sharpening, curves, saturation boost, etc all work to make noise show up more.... try it :)
brightening an image make noise in dark areas pop out where as you may not have seen it before... this is why i do noise reduction last of very very near last....:sport18:

[mickeymoo21]

Hi Michael,

just a bit OT, but what are the advantages of saving in Tiff as opposed to PSD, which is what I have been doing, up until I convert to jpeg before printing/web publishing.? Is there an advantage or is just personal preference?

Im not really sure if there is actually any advantage in useing tiff probably more a personal preference I think, someone who knows more than me may be able to awnser that better though.

As psd is the native format for photoshop (I think?) it may not be as widely supported across a range of other software?? But again I don't really know hopefuly someone will though :D

[mercho]

Alright i have done a bit of testing. This is only basic as i dont know a whole lot about this process to start with...

Done using a combination of CS3 and noise ninja and a very noise source image (jpeg), which i don't know is that good a testing image...

1. Original Section of image 100% crop
http://i28.photobucket.com/albums/c218/mercho/orig100.jpg

2.Noise ninja reduction with 60% sharpening at a radius of 1.2 then CS3 sharpening
http://i28.photobucket.com/albums/c218/mercho/nn_noise_sharp.jpg

3.vice versa of 2.
http://i28.photobucket.com/albums/c218/mercho/nn_sharp_noise.jpg

4.CS3 noise reduction at strength 5 and 25% and de-speckled then CS3 sharpening
http://i28.photobucket.com/albums/c218/mercho/CS3_noise_sharp.jpg

5.vice versa of 4.
http://i28.photobucket.com/albums/c218/mercho/CS3_sharp_noise.jpg


The only major difference i can see with this image, with "basic" editing is that noise ninja has a far better noise removal system over cs3, but the image is a lot softer...

[Lani]

Interesting testing Mercho,
noise ninja seems to a really good job. I might have to try something similar and see what I can come up with as far as processing order goes. :)
Thanks for sharing it.

[peterst6906]

Lani,

There are a few different workflows around, however the general consensus among the recognised workflow experts (Jeff Schewe, Andrew Rodney, Bert Monroy, Eddie Tapp, etc.) is that noise reduction should be done before any sharpening.

Whether or not you complete your tonal adjustments first or second, is a bit more individual, however one commonly used workflow if you have to use dedicated noise reduction software is:

1. RAW conversion (with no noise reduction or sharpening if you are going to use 3rd party software)
2. global tone adjustments (white point, black point, curves, color cast removal)
3. noise reduction
4. capture sharpening (if you use a multi-pass sharpening workflow)
5. local tone adjustments, cloning, compositing, etc.
6. output sharpening
7. printing

Noise reduction and sharpening are very closely related subjects in that sharpening aims to increase contrast along edges, whereas noise reduction aims to reduce contrast in non-edge areas. Noise reduction effectively involves blurring parts of the image and if it has to be applied too strong, fine detail can be blurred away.

If you apply sharpening before noise reduction, you run the risk of sharpening the noise and making the image look much noisier. Then when you apply noise reduction, you have to apply stronger settings to overcome the increased appearance of the noise. This then reverses the effect of the sharpening and also leads to more detail loss.

By conducting noise reduction first, you can reduce the contrast in the large surface areas of the image and make the noise almost invisible to subsequent sharpening. This also allows you to apply less noise reduction and maintain more detail in the image (especially images with a lot of fine detail) and still gives you a good base for subsequent sharpening.

In general though, it's better to apply noise reduction at the RAW conversion stage if you shoot in RAW and then avoid using 3rd party software if possible. This is because all the settings you make in the RAW converter are applied in an ordered sequence designed to produce the best result. Some adjustments are made on the linear RAW data and some on the gamma corrected image data, but in either case, the noise reduction will be applied at the best point in the workflow and you can avoid using 3rd party applications.

In my own workflow with my D200, I use the RAW converter (Lightroom and/or ACR) noise reduction algorithms for anything below ISO800. At ISO 800 and above, I generally then switch over to using 3rd party software.

Hope that helps.

Regards,

Peter

[Lani]

Thank you Peter for such a detailed answer.
To be honest I didn't even know that it could be done at the RAW stage.:o That gives me something to work with. :)

[peterst6906]

Thank you Peter for such a detailed answer.
To be honest I didn't even know that it could be done at the RAW stage.:o That gives me something to work with. :)

Since you are using CS3, most of your noise reduction can be done at the RAW conversion stage and the alogrithms have been improved in recent versions of camera raw.

There are two sliders available in ACR for noise reduction, one for luminance noise and one for color noise. They are located in the Detail Tab, just under the sharpening tools.

As an example, here is an image that I stupidly shot at ISO 800:

http://peterstacey.com/images/forums/noisereduction_1.jpg

At ISO 800 on my D200, I would normally switch from using noise reduction in ACR over to Noise Ninja, but for demonstration purposes, I hope the following is useful.

To access the noise reduction sliders, go to the Detail Tab; and to see the change on the screen, you need to be zoomed into at least 100% magnification (at the bottom left of ACR). You can be zoomed in more, but if you are zoomed out, you won't see the effect on screen:

http://peterstacey.com/images/forums/noisereduction_2.jpg

With this image, you can see that there is a lot of noise. Here is the default in ACR (which applies a 25% color noise reduction as standard):

http://peterstacey.com/images/forums/noisereduction_3.jpg

There is a lot of noise in the image.

The Luminance Slider affects the difference in the brightness between pixels that accours randomly when you take an image (even with a lens cap on, some pixels will register a signal because of noise and this causes differences in brightness or luminance levels). By sliding the luminance slider to 100%, you can see what it does to even out the brightness:

http://peterstacey.com/images/forums/noisereduction_4.jpg

This reduces the luminance noise, but as you can see, it also wipes out detail. So it is obviously best to avoid noise in the first place, but if you can't then slide the luminance slider up until the image looks relatively smooth in the large areas (eg. skies, flat surfaces, etc.), but maintains detail that you want to keep.

Next, the chroma or color noise affects the color of individual pixels). With the slider dropped to 0, you can see the red and green or red and blue coloring that is typical of color noise:

http://peterstacey.com/images/forums/noisereduction_5.jpg

The thing with color noise is that once you find a setting that works for your camera, you can almost take a set and forget approach. Once you find a good setting that is just high enough to remove the color noise, you can apply that one setting to most images (I use a value of 13% for my D200).

So, the net effect on this image of applying very strong luminance noise (70%) and 13% color noise reduction, shows a big difference when you compare the before to the after:

http://peterstacey.com/images/forums/noisereduction_7.jpg

And this is all within the RAW converter, even before you think about using 3rd party software.

Additionally, when I open it in Photoshop the image will be sharper than it appears here, because by applying the noise reduction in the RAW converter, I can also apply capture sharpening there as well and ACR will perform the tasks in an optimised order involving reducing the noise first and then adding the sharpening. This helps to produce the best quality image passed to Photoshop.

Regards,

Peter

[mercho]

snip

Thanks for that info peter! :th3: Appreciate it... I need to start shooting Raw more :o

[ricktas]

I have moved this thread to the tutorials section as I think it deserves to be available long term. Some of the replies and examples are very informative.

Thanks everyone who contributed so far.

[Dylan & Marianne]

I know this is an old thread but I just thought I'd throw another idea out there.
Usually, noise reduction (and sharpening) does not need to apply to the whole image. For instance in a typical seascape scene with rocky foreground and a sky or water that is smooth, the best effect I've found is to sharpen the foreground while reducing grain and noise in the sky/water.
Aside from all the other processing, the 2 methods I've used are:
1.
-Do all of your processing that you otherwise would have done up to the point of sharpening.
-Save this image and apply noise ninja (or whatever other program)
-Save the new image as another file and import it as a new layer on top of what you've done.
- Create a layer mask on this new layer and mask out all of the areas where you want details while keeping the areas that you want to look silky smooth.
- Create a new layer in overlay mode underneath the noise ninja'd layer (and its mask). apply a high pass filter to this layer. Doing this will only sharpen everything that you masked out from the noise ninja'd layer.

#2
- Do everything you otherwise would to the image up to the point of sharpening
- duplicate your background layer (or a copy of the background layer depending on your workflow)
- apply surface blur (radius 5, threshold 5 is what I use for a subtle effect) to this new layer. I find that this removes noise as effectively as the noise reduction function!
- add a layer mask to the blurred layer and mask out the areas which you don't want blurred (eg rock details, boats etc)
- create a new layer in overlay mode underneath the blurred layer and apply a high pass filter.

The results of both of those processes is that you smooth out whatever needs noise reduction without reintroducing noise to the areas that you've smoothed.
It seems complicated but you can actually create an action in CS3/4 that does the high pass filter layer for you , so it's just a 1 click.
Hope that was helpful in some way.