Trouble with Color Noise Correction

[Jeff F]

In chapter 4 of Richard's book, he lists the steps he used to correct the noise in the vince.psd file. Looking at the before and after images in the color pages shows a marked improvement to the most casual observer. The problem is I cannot reproduce his results. I've blurred that color layer like there's no tomorrow, but in the end with the before and after images side-by-side, if I didn't know better I'd think they were identical. He doesn't mention what radius he used in the Gaussian Blur step, so I've tried many combinations to no avail. To my untrained eye, the luminosity layer appears to have about as much noise in it as the color layer after they're split, but there's no mention of a correction to the luminosity layer in the book. Has anyone else tried this unsuccessfully or know what I may be missing?

Thanks in advance,
Jeff

[Richard_Lynch]

You probably will not SEE the same results as appear in the color section with just a single application of the noise reduction. This is partially due to printing in process color on an offset press and comparing it to what you see on your monitor -- monitors tend to be a bit more exacting (and minus one color conversion). At the same time, I am sure I did a few additional things to the images in the color section to show results -- almost a necessary evil in a book

You'll notice at a glance that the after image in the color section is lighter and a bit more saturated...The idea of the color section to me was probably more to show what can be done with several applications of the techniques in the book rather than just showing what occurred in B&W with denoise only. It was at the wrong point in the book to go on about additional corrections and I believe my intent (originally) was to supply the full-blown correction on the CD -- it's quite a tangle keeping all the things I want to get in a book together, and that was an oversight.

If you open the image and do the separation and really look at the Blue channel, then flatten, apply denoise and look again, I think you'll be amazed at the difference. The noise in the Luminosity can be quelled a bit using a curve application (you'll want to drag out those highlights and shadows) and there will be something you'll want to accomplish with Hue/Saturation. You might also use a luminosity mask to mask the Luminosity channel and try to apply some blur in the mid-tone.

I see how that might appear to be cheating, and it wasn't at all the intent -- otherwise I'd never admit it. My goal in the section in Chapter 4 was to work with a concrete example of how Luminosity and color could be used to an advantage -- getting off into the rest of the corrections would have been a tangent. Interesting to some, and valuable, I agree. My goal in the color section, was to give you a good look at what can be accomplished. That leads to some things sliding by in the technique. However, and in defense of the tactic, too much of one image can get pretty boring. For the sake of interest and moving the reader along, I wanted to get through general technique to enable the later, more involved projects. Perhaps in a later version of the book I'll get a few more pages, and this can be expanded.

If you'd like to have a look at the layered correction, I can dig it up...I don't have it on hand right here as this is not my writing workstation. I'd need your email...

Let me know.

[Jeff F]

Thanks Richard,

It's nice to know your purpose for the color section of your book; now I won't get wrapped around the axle trying to get all of my results to mirror yours exactly. I will email you soon with my email address because I would appreciate you sending me the layered copy of the after image when you come across it. When I get into a subject that interests me, there's rarely a such thing as too much information.

Another related question came to me while reading your reply. Is it possible to "Split Luminosity" followed by "Split RGB" on the color layer? In my mind, this would allow you to eliminate the noise in the isolated blue channel without affecting the luminosity one iota, if desired. If this is covered later in the book, just tell me to keep reading. I'm only on Chapter 4, after all.

Thanks again,
Jeff

[Susan S.]

Jeff- by trial and error i've found that I get a better result in chromanoise reduction by blurring the color layer several times using a radius of 2 to 3 pixels - any more than that and the colour seems to start to bleed into places that it isn't wanted. it makes a Big difference over just applying a correction once.

Richard, thanks for the tips on reducing noise in the luminosity channel too, that's useful. Noise reduction is one of my current preoccupations - I was going to buy a better camera, but my gas furnace blew up, and the family weren't that keen on shivering in the cold! Now the idea of splitting the colour layer into RGB and applying denoise to that and remerging sounds interesting. It always does seem like the blue channel that has most of the noise. Jeff, you are asking some most interesting questions....(and getting through the book much more quickly than i did - I must confess that I still haven't worked in detail through all of the examples - i've been sidetracked by other things. I really must get back to it)
Susan S.

[Jeff F]

Susan,

Thanks for chiming in with the 2-3 pixel radius technique on the noise reduction. I played around with it last night for a little while trying to reproduce Richard's results. However, I was operating under the apparently false assumption that one full blown correction was better (or at least less lossy) than several smaller adjustments. I will definately try it your way now.

To be honest, I'm working through the book examples in depth purely as a way to understand it (inside out if possible). Other than some very basic levels adjustment and some red-eye correction, I've not been seriously into photo correction before now. But now with the more pictures I capture with my digital, I getting to be more discrimminating and critical of how they look raw. Anyway, I want to learn as much as a hobbyist can about this field.

By the way, it's been over 30 degrees C here in the Mojave Desert all week, so bring the family up here and buy that new camera -- I promise there'll be no shivering!

Thanks,
Jeff

[Susan S.]

Jeff - you pay the airfare and I'll be there!! (I'm in Australia - and it was 20 degrees C today, so i can't complain too much, although three degrees C first thing this morning was a bit nippy!). If I lived in the US I would already have bought the new camera - I have to pay between thirty and fifty per cent extra on US prices here. (but i prefer to have a warranty so I'm not game to import)

I often find, particularly with adjustments that require blurring that several small increments can be better than a one shot large blur. Obviously the less blurring the better, as it's all destroying data from the original image. You certainly do need to check and compare with the original. (and remember to work on a duplicate! - I've been caught a couple of times by that, but haven't destroyed anthing valuable).
Susan S.

[Richard_Lynch]

Is it possible to "Split Luminosity" followed by "Split RGB" on the color layer?

Jeff, try the RGBL separation. If you are new to this and these are the questions you are asking, you'll pick it up pretty quickly. The idea of RGBL is pretty new, and as you say (I believe it does in the book too) offers the advantage of using RGB for color and luminosity for tone control.

Susan, using the blend mask techniques to isolate specific areas where noise occurs, you can really target exactly those tonal ranges where the problems occur.

If people bought big books, I'd have written one covering all this stuff more extensively! I'm glad to see this kind of question here.

[Susan S.]

Thnks Richard. I've been working in that direction. While I can improve matters I'm rapidly coming to the conclusion that I really do need a new camera!! (and I buy thick books, when I can get them - if they are any good that is - but a lot of the doorstop - size books that I've looked at seem to be best off used for that purpose..)

The RGBL separation is one I haven't played with yet - but I will now. I was trying to get my head around the way it works. At first sight I couldn't see that the luminosity layer would add anything to the mix, as the three colour layers clearly add up to the total image already, but it obviously does as otherwise you wouldn't have put it in there! ... Now it makes a bit more sense as it does indeed seem to act exactly like a split of the colour layers from a luminosity/colour separation. I'm still a bit befuzzled as to why it works like that - I still haven't quite got straight the interaction of multiple layers in different blending modes. When I've got just two layers I can usually predict what is going to happen, but multiple layers like that are still a bit tough.

Susan S.

[Stephen M]

These resources are for the full version, but users of PSE may find some benefit to them...

* For more on LAB mode in general (luminance and color blend modes are similar but not the exactly the same):

http://members.ozemail.com.au/~binaryfx/PSTV_links.html#L

* Luminance masks or layer blend data:

http://members.ozemail.com.au/~binaryfx/ST_utilitymasks.html

* On colour component noise cleaning:

http://members.ozemail.com.au/~binaryfx/IT_artifactremoval.html

* Grain and noise info:

http://members.ozemail.com.au/~binaryfx/PSTV_links.html#G

* For more on layer blending, see these RP forum link:

http://www.retouchpro.com/forums/showthread.php?s=&threadid=4492

Sincerely,

Stephen Marsh.

[Jeff F]

Richard,

Thanks again for the RGBL information. I just finished Chapter 4 last night, and you mention it in there near the end. I just need to be more patient it seems. I haven't had a chance to play with it yet, but rest assured I will. It's too bad your publisher won't let you do abridged and unabridged versions of the book. If Amazon charged me shipping by the pound for the books I've bought, I couldn't afford them. Until the long version comes out, I'll be here trying to pick your brain.

Stephen,

I appreciate the links you replied with. They look to be chocked full of good information.

Susan,

I'm with you on the whole "multiple layers with different blending modes" business. I really have to stop and carefully work my way through, and many times I have to go back and look up what exactly the blending modes do because I haven't really worked with them much. Oh, and if I had the kind of money they charge for airfare between Australia and the US, I'd just send it to you and you could probably get several nice cameras plus the full version of Photoshop.

Thanks to all,
Jeff

[Richard_Lynch]

Jeff and Susan,

This reply is directed perhaps not so much at you as in that direction...Hope it helps a little.

Controlling layers and layer modes as opposed to playing with them takes time to understand. You need to digest color theory and understand what the layers do to one another, not necessarily just what the mode means. I've listed the descriptions in the tool menus and in the back of the book so you have access to the descriptions -- that only gets you part of the way there.

The difficulty in trying to teach a Photoshop or Elements 'method' is that people want fantastic things to happen. They often go right for the filters...and Layer modes (used on the tools as well) are not far behind. My digital art students play with these things all the time -- not really understanding what is goig on, but enjoying the effects. The misunderstanding gained is that these are toys and incomprehensible. Methods are pretty much boring theory until they can be applied to make something fantastic happen.

Even with a lot of experience trying filters and modes, you'll see I only really use a handfull of filters in the book (noise, blur, sharpen, clouds) and only a few modes (Luminosity, Color, Screen, Multiply, Overlay). As i 'preach' at the beginning of the book, there are a few tools you will learn to use all the time -- and those tools will have most used modes and functions. Focusing on these will give you fewer things to remember. But these modes are really fairly simple in function, and that is why they really become staple: they are easy enough to figure out to apply with consistency (and overlay may even stretch that definition). Part of the problem may be that you are trying to think of the multiple layers cumulatively. More important is to consider the layer groups -- as that is where you are really doing complicated calculations. To over-simplify: if you have a group wher the botton layer is a Screen mode layer and that stack is topped by 100% green (0,255,0) Multiply mode layer, it will be effecting the green channel -- no matter what or how many changes and undercurrents fall in the group. Unless you have another such group (and I'd try not to do that unless you are quite conscious of why you are doing it), that group is, in effect, your green channel. Think of it separately from the blue and red, and the mass of calculations becomes less challenging.

there are other ways to accomplish what I was doing in the book, and some are damned near magic -- they'd have taught you nothing about color and mode and what is affecting the image. I chose what was sometimes a little more difficult because it makes a little more sense. It is also a book that gathers YEARS of learning about digital images. If you can pick it up and in months or weeks understand even half of it as well as I, not to boast, but I've quite done the job I set out to do -- and then some. If you are picking it up quickly, all the better -- but don't get fooled into thinking it is easy, even if you are a quick study (you'll see a few reviews on amazon that are result of thinking even the advanced is 'easy'). I've tried to show results in the book that are easy to accomplish while demonstrating the power. That CMYK separation...if you really get it, you are beyond 98% of users -- Photoshop or otherwise. It is a complex assembly of cumulative theory in the book, and practice that spans pre-press from before there were digital images and how pre-digital separations were made using cameras and film. There are digital masters who would say you couldn't do it. it isn't their fault, they've never really thought to try because Photoshop will do it for you -- leaving them without a clue as to how it works or why, and putting the results in the hands of calculations made by Adobe engineers.

All this is to say: take your time and absorb it. Keep the mode definitions on hand. See an image and try to figure what you want to do with it and how before you go to it...and if it fails, figure out why it didn't work. That is learning from mistakes, and it will lead to the permanent understanding of one mode at a time.

OK?

[Susan S.]

That's interesting Richard. Thinking about the layers cumulatively is exactly what I was trying to do. I will readjust my thinking. And it is the CMYK chapter that I've got bogged down in....but I'd rather have it more difficult if in the end I get a better understanding of how photo editing works. (trouble is, at the moment, for that particular chapter that end is not yet in sight - reflecting the fact that I really need to go back to the earlier stuff and go over again things that I thought I understoood, but obviously didn't quite). Part IV was much more straightforward, techniques that I knew bits of or variants of anyway, and it just put everything nicely into place. Part V I've only dipped into.

Susan S.

[Jeff F]

Richard,

I, too, like Susan would rather it be difficult if I will garner a better understanding in the end. There's a world of difference between knowledge, understanding, and experience. I know and understand very well how a jet engine works, but I'd be hard pressed to build one right now. My knowledge and understanding of the color and layer concepts are skyrocketing right now thanks to your book. But I know it's going to be quite some time before I can start to apply them to my own photos reliably and with good results. For example, the "bottles" picture in your book in which we isolate the blue bottles so as not to make the same color changes that we make to the rest of the picture: I can do all of the steps now and totally understand why it works the way it does. But knowing this procedure, if you were to hand me a picture that needed a similar fix, I wouldn't make the connection. It's very difficult for me to see what a picture needs to look better. I can only hope that it will come with experience. Until then, I'll plug along increasing my knowledge and (hopefully) understanding and wait for the experience to catch up. I haven't gotten to the CMYK separations section yet, but it looks like I'll be in for a treat.

Thanks for all of your help,
Jeff

[Richard_Lynch]

Jeff,

The art of change is not something I really address -- and I think it is a whole other topic. I guess my point would be to give you the controls so that you can test out which things you think you want to change, powered by the understanding of how to do it. I also can't guarantee a right and wrong in an artistic decision: if you wanted to make the bottles another color, it is possible it would be better than the color I chose -- either as a general concensus or in your own eye. I think an appropriate name for *that* book would be the Art of Photoshop (I am not advocating anything that may exist by that title). I hope, however, that you are telling me you'd be able to change a specific color, but just wouldn't know which to choose. You'll see later in the book that I try to make sure I am not suggesting my corrections are the only possibilities...For example, changing and recreating the scene in chapter 7 is a judgement call. You may either like the result or not, and that isn't the point. The point is to learn the technique of re-defining, creating possiblities, and giving you the tools to apply.

Eventually you will get to the point where you look at an image and make a list of what you want to change...Playing with images and seeing the possibilities (and how to carry them out technically) is a first step in the process.

Susan,

it is fine to go back...less good to skip forward...and probably possible to go entirely through without absorbing a thing. The latter being possible but not desireable from my perspective. My suggestion would be that going slowly and repeating sections before moving on will give you the greatest understanding of the result and the best possibility of making appication successfully to any image.

Eventually you will both get it all...I have no doubt.

[Jeff F]

Well, if you ever decide to write such a book (on the art side) I'll probably be first in line. There's no question that the artistic side of my brain has atrophied from disuse. I'm making photography my one big endeavor into the artistic realm if for no other reason than to preserve memories, which are important to me. The saddest thing is that many of your "before" photos look just great to me until I see the "after" version. Then I look back and see that maybe the "before" image wasn't so great after all. But why didn't I see it the first time? Lack of experience? Lack of discriminating taste? My daughter is very artistic and has been all of her life. That makes me wonder if it's something one must be born with, and I'm just out of luck. I love to sit at the computer and play with an image to try and make it look its best, but I don't have all the time in the world to sit and do random corrections with every picture I take and hope I eventually find the right combination. After all as you say, there is more than one way to go at it, and, therefore, thousands or more combinations of ways to change a photo. Therefore, I need to learn to be able to see what it needs up front to cut down that time. That, I'm afraid, will be a long, long journey. On the other hand, there's something to be said for a long journey as long as you enjoy yourself, I suppose. I've probably rambled on here a little too much, but thanks for your time.

Jeff