Go Back   Steve's Digicams Forums > Digicam Help > Newbie Help

Reply
 
Thread Tools Search this Thread
Old Nov 29, 2005, 1:31 PM   #21
Senior Member
 
Join Date: Oct 2003
Location: Indian Rocks Beach, FL
Posts: 4,036
Default

Quote:
Contentious opinion, Slipe! Brave man...
It wasn't meant to be contentious. I just do things in a different order which makes sense within my concepts of what I am trying to accomplish.

I'll give you an example of the noise reduction. This is a picture I grabbed from the first folder I opened. Suppose I wanted to print this picture and for whatever reason wanted the background to show up. It is the opposite treatment you would want for this particular photo, and other than paste in some open eyes from another photo it isn't one you would print in the first place. But I find I often want details in shadows brought out and this shows some of the problems as it relates to noise reduction.

This is the snapshot:


This is a bag in the background:


Same bag with a default dose of Shadow/Highlight:


Some things that are obvious to me:
Considering the bag is in the background and a small crop, I would not apply noise reduction to print the original with the dark background.
If it did apply noise reduction to the dark background it would be subtle.
If I printed the photo with the background brought up I would require a lot of noise reduction for the background.
I would apply different amounts of noise reduction for the background I brought up, the munchkin's face and the teen's face that needs a little extra for the complexion and also doesn't want the edges of the blemishes eliminated from the noise reduction. That is all part of normal workflow and there is no way I could apply noise reduction as a first step. I would have no idea where and how much I needed.

Most "experts" who write Photoshop books and tutorials say that sharpening should be the last step. That is generally true, but not necessarily with properly applied noise reduction. If you find the edges and invert before applying noise reduction you aren't effecting your sharpening much because you are avoiding edges. But sharpening accentuates noise and should be a consideration in your noise reduction. So I sharpen first if I am going to do a sophisticated noise reduction. Since sharpening should be a last step it follows that if you elect to apply a noise reduction with the edges excluded after sharpening, then the last step should be noise reduction. I do sometimes tweak the sharpening as a last step, but it isn't usually necessary.

The only way doing an upsample after noise reduction would degrade the noise reduction would be if you had a profile made for that particular amount of upsample. Except for that unique case, you would want to apply the profile to the original to get the proper noise reduction from the profile. Doing an upsample first would degrade the profile. If you don't use a profile I don't see it makes any difference other than the noise reduction takes a lot longer after the upsample. Everything else takes longer after an upsample and before a crop. I don't think the end result is different, but it is pretty inefficient to upsample early and crop last. Cropping before noise reduction could possibly degrade a profile – I don't know as I don't usually use profiles.

Norm in Fujino I think those quotes presume working from a pre-made profile. I think they consider that a normal mode of use for Noise Ninja and Neat Image. If you make a profile for a particular camera at a given ISO for certain conditions, then applying noise reduction is a lot faster. Most post processing and especially sharpening would negate the value of the profile, which is usually made from the image as it comes from the camera.

I usually shoot "soft" with minimum sharpening and contrast to get maximum dynamic range and minimum artifacts for post processing. If I apply sharpening and contrast (levels, curves etc) after running a noise profile I find I still have to do a noise touch-up and have to make a custom profile anyway. The same holds true if I select certain areas for noise reduction – no profile would be much help. Since noise reduction isn't a standard process for me and is only applied in limited situations, I don't usually apply profiles.

I do agree that if you are using a pre-made noise profile it has to be applied before most post processing. But if you are making a custom profile for a particular image or part of one, you can make better judgments after all of the other processing is done.


slipe is offline   Reply With Quote
Old Nov 30, 2005, 4:44 AM   #22
Junior Member
 
Join Date: Nov 2005
Posts: 24
Default

This has turned out to be a steep learning-curve... I did think that there might be a 'right' way of doing this, after all, it's just maths - brighten this dot in relation to the ones next to it if it's so bright, and such and such a colour; put a new dot here with colour and brightness halfway between it's neighbour on this side and it's neighbour on the other side... I kind of thought that I might just be able to apply a set of rules which everyone agrees on. That's what maths is about.How wrong was that?!

But I have to say, I've picked up a lot - I have some books to look out for (thanks, jlacasci) and a changed perspective on things - It hadn't really occurred to me that some of the things I had always thought of as applying to the whole image might actually be not a great deal different from any other 'effect' - like sharpening, for example: I just hadn't considered that I might sharpen one part of an image more (or less) than another part...

Regarding the noise-reduction, I guess I will have to experiment a bit. At the moment, I use Neat Image to do the whole image in one go. I'm leaning towards doingthis first (thanks, Norm, Caelum) if the idea is to end up with a slightly-processed recognisable photo. I'm thinking that if I was going to do a paint-effect, I might not do noise-removal at all - maybe the noise will add some extra texture...? And if I get a photo which needs different treatments to be applied to different areas, then I'm going to need to refer to this thread again, because I can't remember waht people have said!

I think I've found a good forum, here...


shambles is offline   Reply With Quote
Old Dec 1, 2005, 3:14 PM   #23
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

I wrote this long follow up to Slipe's comments, but lost it when my browser crashed. I guess I'll do what I can do cover some of it.

Slipe,
In my opinion, sharpening is a non-linear process. It does not apply the same thing to everywhere in the image. Therefor the noise is effected in different ways in different areas of the image. One of the founding assumptions of noise reduction is that the pattern & characteristics of the noise is the same everywhere in the image. After you sharpen it is my believe that this is no longer true. This chain of logic (maybe flawed, but I don't see where) makes me never sharpen before I use noise reduction software.

I can't don't believe it does as good a job if at removing noise if I sharpen first. Simple tests I've done have born that out, but they weren't exaustive tests. NI can do sharpening and noise reduction in the same operation. I often do that. But I wouldn't do USM before noise reduction. Even if I train it on that image (which is really the only way I do it.)

I don't use canned profiles or noise ninja. But those quote absolutely are true if you use a canned profile. But I think they are also true if you don't. As per my reasoning above.

shambles,
To keep your math analogy going, many processes in editing an image are commutative so it doesn't matter what order you do them in. Change the contrast and then lighten the image... or reverse it. The results will be the same. But there are a few things that are not commutative. It is my believe that sharpening is one of those things.

So in general you'll find a pattern you like to work in. That is what I suggest.

I agree with you that leaving the noise can add texture. If that is the look you want, go for it!

You will find that many images will benefit from localized adjustments. Lighten and sharpen the eyes more to draw attention to them. Things like that.

Well, that is some of what I wrote before.

Eric
eric s is offline   Reply With Quote
Old Dec 4, 2005, 12:41 PM   #24
Senior Member
 
Join Date: Oct 2003
Location: Indian Rocks Beach, FL
Posts: 4,036
Default

Quote:
Slipe,
In my opinion, sharpening is a non-linear process. It does not apply the same thing to everywhere in the image. Therefor the noise is effected in different ways in different areas of the image. One of the founding assumptions of noise reduction is that the pattern & characteristics of the noise is the same everywhere in the image. After you sharpen it is my believe that this is no longer true. This chain of logic (maybe flawed, but I don't see where) makes me never sharpen before I use noise reduction software.

I can't don't believe it does as good a job if at removing noise if I sharpen first. Simple tests I've done have born that out, but they weren't exaustive tests. NI can do sharpening and noise reduction in the same operation. I often do that. But I wouldn't do USM before noise reduction. Even if I train it on that image (which is really the only way I do it.)

I don't use canned profiles or noise ninja. But those quote absolutely are true if you use a canned profile. But I think they are also true if you don't. As per my reasoning above.
Hi Eric: I agree with everything you said. I don't see how anything in the above quote disagrees or has to do with anything I have posted in this thread though.

I sharpen last if I apply noise reduction to an entire image. I also sharpen last if I select out my edges and apply noise reduction to the entire image. It is a well established rule of image editing that sharpening should be the last step. And I agree that sharpening and perhaps other editing would interfere with a full image noise reduction.

The only exception to that rule for me is if I have particular parts of an image I want to apply a separate dose of noise reduction to. That usually occurs if I bring up the shadows because of a dynamic range problem, poor exposure, or a flash shot out of range or with a dark background I need enhanced. It seems apparent to me that I can't identify those areas before bringing up the areas with contrast masking, shadow/highlight or selective levels/curves. So there is no way I could make those applications as a first step.

Another reason for selective noise reduction is faces. Teens often have complexion problems that noise reduction helps with. Older people often look better with wrinkles softened. And sometimes a face just looks better softened a little. Faces have to be handled differently from other noise reduction IMO. You don't want to mask out edges before applying the noise reduction or the blemishes etc are still sharp. Since almost by definition such noise reduction is reducing your sharpness you wouldn't want to apply that to the entire image. I use layers and the eraser or history brush to keep things like eyes sharp in the face. But I can't get the face right without doing my overall sharpening first.

In many cases where specific areas need noise reduction the rest of the image is better without it. I haven't used Neat Image for a while. I realized I needed a plug-in and someone gave me Grain Surgery. Maybe Neat Image has improved to the degree that noise reduction doesn't affect image sharpness at all. Or maybe an overall application as a first step at a level that doesn't degrade the image will make deep shadows brought out noise free. That wasn't the case when I used it. Grain Surgery certainly isn't that sophisticated.

After selecting out the area that needs special treatment and masking the edges, smart blur works almost as well as noise reduction. Maybe I should look on my selective noise reduction as a blur step. But sharpening affects the amount of noise reduction I have to apply in most cases, and the step works better if I sharpen first.

I have seen many of your images and they are usually better than anything that comes from my cameras. I don't doubt that you are sufficiently talented and your DLSR sufficiently competent that your main use for noise reduction is a general application for high ISO shots. Neither my talents nor cameras allow me the luxury of not having to give special treatment to specific areas.

slipe is offline   Reply With Quote
Old Dec 4, 2005, 1:50 PM   #25
Senior Member
 
Join Date: Jul 2003
Posts: 3,396
Default

Slipe, I hope you don't mind.
I think your sample image is rather nice!


I retouched it a tad (about 5 min work with the new stuff I am learning lately on prepping images)

Last edited by PeterP; Apr 17, 2013 at 11:43 PM.
PeterP is offline   Reply With Quote
Old Dec 4, 2005, 4:21 PM   #26
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

Clearly I missunderstood your post. I agree with everything that you agree with.

I always edited with layers (adjustment whenever possible) in PhotoShop CS2. If I were to find noise in the dark parts, I would duplicate the bottom layer, apply noise reduction to that, apply a mask to that layer and then undo some of the noise reduction via the mask (like on a face.)

Now, if I found the extra noise in the darks because I had applied a contrast adjustment via Curves or Levels, I'm still in a good situation. If I do what I describe above, I'm still working on the unmodified bits of the background layer. So eventhough I saw the noise because it was made more obvious via the Curves or Levels adjustment, since they are on a separate layer I have the flexability to apply the noise to the unmodified layer.

Does that make sense? So if you find the noise is too much after you see it (because the contrast/brightness adjustment brought it out) you can still do noise reduction to the unaltered bits for best performance.

And thanks for the complement on the quality of my pics. I don't know what to say....

You've mentioned "selecting the edges" several times how. How do you do that? NI still removes a little sharpeness, but it has a sharpening setting which puts it right back in (and some times produces an even sharper image afterwards.) So that is what I do. But masking out the edges sounds like an interesting technique.

Eric
eric s is offline   Reply With Quote
Old Dec 4, 2005, 11:57 PM   #27
Senior Member
 
Caelum's Avatar
 
Join Date: Sep 2005
Posts: 1,030
Default

PeterP wrote:
Quote:
[...]I retouched it a tad (about 5 min work with the new stuff I am learning lately on prepping images)
I'm really sorry to say thisPeterP but while I find the original charming I find your touched up version a little disturbing. I think the trick to touching up is subtlety. Myfeeling is thatthe natural humans faces have been turned into plastic masks.
Caelum is offline   Reply With Quote
Old Dec 5, 2005, 6:00 AM   #28
Senior Member
 
Join Date: Aug 2002
Posts: 851
Default

Shambles, I strongly suggest you get an Uninteruptable Power Supply (UPS). Then those brief power fluctuations will not crash your computer, it will breeze right through them, not even knowing they happened. In the event of a major power failure, you will have time to do an orderly shut down of your system, saving your work first.

To speed up your photoshop editing, make sure you have shut down all other programs and especially all that junk that starts automatically when XP starts up.

Next, assign your page file for PS to another physical drive than the drive your OS is on. That should improve performance a little. Finally, consider adding more RAM.

Declan
amazingthailand is offline   Reply With Quote
Old Dec 5, 2005, 8:05 AM   #29
Junior Member
 
Join Date: Nov 2005
Posts: 24
Default

UPS is on my wishlist. One day... Hopefully, if girlfriend has taken my suggestion on board, I might get a yummy new SATA HDD for christmas!

Kind of staying with the theme of this thread, I tried to download Noise Ninja to do a comparison with Neat Image. According to one of my security programs (Spybot S+D), Noise Ninja contains a somewhat dodgy component called 'Winfixer', and as a result of this warning I aborted the install. Has anyone else had this experience? I can't decide if it is a false positive or not and I don't have the skills to unpack the installer to check. Also, someone earlier in the thread mentioned DxO optics. Interesting stuff. I found a program called 'PT lens', which appears to do similar stuff around correcting for lens aberration. I've installed it (Manually, there are two or three bits which need to be placed in specific system folders) and it looks like it works, but until girlfriend gets her new christmas present, it won't be possible for me to check it out properly. Has anyone had any luck with that program?
To get back on track, I suspect that correcting the image for lens-created anomalies might come after general noise-reduction? Unless of course, the noise reduction is being done to different levels for different parts of the image?
I'm struggling a bit - this is just way more complicated an answer than I expected...
shambles is offline   Reply With Quote
Old Dec 5, 2005, 11:06 AM   #30
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

Good luck on the SATA drive. There are some very good ones out there.

If you feel that PT Lens alters the image in an inconsistant way across the image (different effects to different parts) then you should definitely do it after noise reduction. One of the fundamental assumptions about noise reduction software is that the noise in a specific portion of the image is indicitive of the noise across the whole image. So you train it on a specific portion of the image (an area with no detail, like a sky or the side of a building) and then it uses what it lears from that portion to remove the noise from the rest of the image. If you chance the noise in some way (like sharpening) then it that fundamental assumption is broken and it won't be able to remove the noise very well (and in fact may make the image worse.)

Image editing is more an art than a science. You look at the image, think about what it needs (more contrast, maybe some satuation, maybe the greens look flat and need enhancing) and then start editing it. You'll get into a habit of what order to work in. There are a few things that have an order (The points we've been going back and forth on in this post) but in general many things can be done in many orders.

If I were to do different amounts of noise reduction in different areas of the image I would still all do them in order, back-to-back and then do the rest of my adjustments. Slipe has a point, if you don't see the noise (the dark area is too dark) how do you know you'll want to remove it? My answer if with adjustment layers in PhotoShop CS2. If I see the noise in the dark area, then I do the noise reduction on a lower layer, one where the dark area hasn't been adjusted yet.

Eric


eric s is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 12:48 PM.