Color Noise

[Jeff Bellune]

Lacking experience with more advanced image manipulations, I am beginning to rely on HP's methodology to solve certain image problems.

I also know that there are usually eleventy-seven ways to accomplish a task inside of Elements.

My question centers on correcting color noise using Split RGB versus Split Luminosity. I have an image with obvious noise in the blue channel. Red and Green are virtually noise-free. I ascertained this using Split RGB and Preview, and I attempted to correct with Split Luminosity. The results were unremarkable. I could see no significant change to the image until the color was blurred so much that it introduced a color shift in the flattened image. On a hunch, I took the flattened image and separated it again into channels using the Split RGB and Preview. I then applied a Gaussian Blur to the Blue channel only until the visible noise was gone. I flattened the image again. I got much better results using this method for this particular image. I was surprised that the blurring of the Blue channel didn't affect the crispness of the overall image very much at all.

Am I imagining this? Or is this indeed a valid method for reducing color noise in certain images?

Jeff

[Stephen M]

Jeff, your post about HP confuses me - what features are you using? I personally stay away from HP for imaging, but they do make some good prints...but this is just my personal bias.

Does Elements offer the ability to set a layer or fade an edit to COLOR blend mode (hue & saturation with no luminance)? If so try a large blur set to color blend to remove colour noise.

Yes, you can directly attack the blue channel - but it is not the best way. You get better results with indirect filtering via color blend or in LAB or other spaces which separate colour from tone.

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

Regards,

Stephen Marsh.

[Richard_Lynch]

Jeff,

I think that it is very much a valid means of correction in certain images and you are probably NOT imagining the result. Image corrections will always depend more on the image than the technique. You might often use a paper clip to hold paper together, and it is best used for that, but you might sometimes find it works for removing discs that are stuck in a computer too (if you've never done it, that little pinhole next to your CD drive fits an un-bent paper clip and can eject either the tray or the cd).

The idea of giving you more than one type of separation is that all of them become important at one time or another. Again, in 300 pages in a book it is IMPOSSIBLE to mention every use of a particular tool -- that is, without writing a book just on that particular tool. That book would get pretty, um, boring. You might use the red channel for creating infrared images (or as the basis of), the green for converting to B&W, and the blue for masking or other support, etc.. While I might suggest a particular tool for most cases (my luminosity-based noise reduction works best for low light digital photography, and images with pretty good resolution), that doesn't mean it is best for every situation, that there won't be other solutions (as you suggest), or that particular images will never require other techniques. In fact, I believe I suggest you use things together. Part of the reason for the more advanced RGBL correction is to take the Luminosity and Color correction to another level. You may be to the point in your explorations where looking more at this will yeild some advantages.

Things that would affect the result lie in the image in question. If it is a scan of an existing photo or if it is a digital image, that quality of capture will be different. The lighting in the scene may be telling of the result, as might the subject.

So, as long as you are getting the idea that separating components can yeild opportunity for specific correction, you are getting both the point and the benefit of the tools.

Stephen M.,
HP is a reference to Hidden Power, not Hewlett-Packard. You might be interested in checking out the book and website: http://hiddenelements.com.

[Susan S.]

Stephen: In this part of the forum HP refers to Richard Lynch's book rather than a well known manufacturer of printers! If I understand your full-version-of-photoshop-talk correctly, what Jeff tried was the Elements equivalent of shifting to Lab mode and blurring the colour component - it requires rather a lot of steps but Richard has provided some actions which allow us poor cousins to do it in one click.

Susan S.

Rats! Richard got there while I was posting!

[Jeff Bellune]

Thanks!

Stephen, Richard, Susan,

Thanks for the replies. As I get more experience working with the basic elements (no pun!) of digital images, I should gain more confidence in things I try when the result looks like I want it to. I still worry that I may not be using the "recommended" tool correctly.

Jeff

[Richard_Lynch]

Jeff,

While I know what I use my tools for, I don't know that I have the only use, nor that I can decipher every possible creative use for every tool. I think it is best to experiment. Every time you think "it probably wouldn't work to..." Do it if you aren't sure. You can always undo (be sure you have enough History levels first!). You may discover something. It may be something other people already knew that you had to learn the hard way or it could be something new. In either case you learn. When you learn, that's when you are in control and use the tools, rather than the other way-round.

[Stephen M]

<BG> What can I say? Graphics has way too many acronyms! <g>

Glad I gave you all a chuckle - myself included, now that I know what Jeff was talking about.

So, my suggestion was what was what was being done in the first place. Oh well.

I would suggest that this edit be isolated and not performed as a global one.

Yes, directly attacking a problem channel may be required - but IMHO, not as the first step. I first use indirect colour filtering before going into luminance filtering of one offending channel.


Stephen Marsh.

[Richard_Lynch]

Stephen,

Believe me, I saw the HP abbreviation early on as a potential problem, and even tried to use HPT, but I don't think the book is just about the tools -- AND I was over-ruled by other posts.

Very good point about not making that correction first or without some other color balancing...Those with the book should know that from the correction steps that are outlined.