Green and Magenta Bands

[SunderJ]

Good afternoon - this is my first post but hopefully I'm not asking a question that has been done to death in this forum (I haven't seen the exact issue discussed so far).

In several of my shots (such as the one attached), particularly when I have to make large changes to exposure in skies, I notice that green and magenta "bands" are introduced.

If the problem is obvious then read no further, however, I have provided below a summary of my attempts to diagnose/correct thus far:

1. I have developed from jpeg and raw files

2. Processed in photoshop in 16 bit pro-photo as well as in lightroom (which I believe is the same).

3. In Michael and Jeff's  LR2 video they note that color noise is often seen as green and magenta splotches - I think this may be a related but different issue (as the subject image was shot at ISO 160 and no amount of color noise reduction has any effect

4. The original image was "Exposed to the right" and highlights were not lost, so I don't believe this has anything to do with the banding that is sometimes seen when pushing darker tones (captured with fewer tonal values) into a brighter space.

5.  Tried to use Photoshop "Select color range" and then flip into lab mode to reverse the offending color through a new selectively applied filter, however, this didn't work. I had difficulty properly selecting the appropriate range given how subtle the coloration is.

6. For reference, I am working on a calibrated monitor which does not appear to be introducing the error (I have seen it while viewing on other monitors as well). Also, I have run off prints (Chromira) and while the offending coloration is less noticeable, it is certainly still present (and therefore not likely a monitor issue)

7. I have seen this on several shots, and while the exact shape and way in which it manifests itself, it is still there. Have seen it on a handful of images, ranging from those shot on my P&S to the D90.

8. I may be wrong, but I almost thought that I saw the same issue on some of the images that Jeff and Michael were working on in the LR2 tutorial, though I don't believe the specific point was ever specifically addressed.

Appreciate any insight anyone may have.

Best,

Sunder

PS I don't know if the photo attachment is working so I've also included a link to see a small version of the image (hosted elsewhere) :

http://www.sunderj.com/Travel/California/C...760_uWWMK-L.jpg

[Paul Sumi]

I'm no PS expert, but I sometimes see this when I set a Curves adjustment layer too radically.  If you are trying to lighten the darker tones, you might try the Shadow and Highlight tool and then use the Curves adjustment layer.

Paul

[Schewe]

Appreciate any insight anyone may have.

The main question is: do you do any color adjustments to the grayscale image _BEFORE_ doing substantial tone curve adjustments? If so, that's why you are seeing what is in effect, color banding being introduced...(and no, it has noting to do with color noise).

Don't know how you are getting to the B&W conversion either...Lightroom/ACR and Photoshop have some pretty good tools but the ideal would be to do the color>B&W conversion as far upstream as possible and then do all the toning you want for curving in the resulting grayscale file. Only when you have your final tone would you then want to convert back to RGB and add in the sepia color tones.

You can prolly fix your existing file by converting to B&W, then back to RGB and re-add the color tone. Note you may also need to add some very gentle grain (unequal noise channel by channel) to help get rid of some of the banding...but at least after the B&W conversion, the banding won't be green/magenta anymore.

[smahn]

Don't know how you are getting to the B&W conversion either...Lightroom/ACR and Photoshop have some pretty good tools but the ideal would be to do the color>B&W conversion as far upstream as possible and then do all the toning you want for curving in the resulting grayscale file. Only when you have your final tone would you then want to convert back to RGB and add in the sepia color tones.

Jeff, does your response assume he's doing the brunt of the work in PS?

He may well be, but the reason I ask is because I've heard it said that LR doesn't care what order you make adjustments, it will always conduct them in the best possible order.

So, what's the deal if he's doing all the work in LR? Again, he may not be, but he said he thinks he's seen the same thing happen in LR demos.

[Schewe]

Jeff, does your response assume he's doing the brunt of the work in PS?

Yep...the way he wrote his problem up, it sounded lie a grayscale>color>mono Chromatic banding issue showing as a result of Photoshop editing, not ACR/LR editing...

[SunderJ]

Thanks for all the "non-expert" and "expert" responses. I have been working through what you have suggested and also wanted to add some clarification.

Jeff, ordinarily if I am doing a BW conversion I first use channel mixer in monochrome as a layer, and then then add curves/etc.

However, in this case, I did it all through LR2 using a preset as a starting point and then tweaking curves/grayscale mix/etc. , so I'm not actually sure at what point the conversion is taking place. I assumed, as others have pointed out, that it does not matter in LR2.

I did try bringing the image out to photoshop, adding a channel mixer layer to put it into BW, and then going back to LR2 to then add the toning (again using the mixer) , however, the same magenta/green issue re-surfaced (albeit a bit less notably).

I'm a little confused by the relation of the BW conversion with the color bands, however, as I have seen this issue in full color images as well.

I attached an example shot, which is not fully processed yet (so excuse the sloppiness), but has already started to exhibit this problem as I darken the clouds.

Lastly, is there a tutorial somewhere demonstrating the technique of adding unequal noise channel by channel as you described? I have not seen this before.

Sunder

[Panopeeper]

Sunder,

in this particular image the WB is incorrect, that causes the strange colors on the clouds. The sky's color is unnatural, the water is strange, but, most importantly, the clouds are reddish. A sample from the very brightest spot of the clouds shows (227,220,212); it should be the other way around, the blue the highest. (The ship is not suitable for picking WB, but there is a small sail on land above the bow of the ship, I guess that can be taken as white.)

[SunderJ]

Sunder,

in this particular image the WB is incorrect, that causes the strange colors on the clouds. The sky's color is unnatural, the water is strange, but, most importantly, the clouds are reddish. A sample from the very brightest spot of the clouds shows (227,220,212); it should be the other way around, the blue the highest. (The ship is not suitable for picking WB, but there is a small sail on land above the bow of the ship, I guess that can be taken as white.)

If I understand your comment, would the natural solution be to adjust the white balance cooler? I agree there is a warm/yellow/reddish cast on the image, particularly noticeable in the sky/clouds.  However, even when I adjust the WB cooler, the green/magenta areas remain (Albeit less noticeably).

Is white balance the only tool I should be using to correct this?

[Panopeeper]

If I understand your comment, would the natural solution be to adjust the white balance cooler?

Honestly, I have no idea if there is any connection. However, I think the best approach is to eliminate other issues first - in this case the WB - so that one can be sure, that the problem is not from that corner. So, don't be surprized if the banding remains there after the WB correction.

Btw, are these shots from the D90? (There is no Exif in the JPEGs.)

[SunderJ]

Honestly, I have no idea if there is any connection. However, I think the best approach is to eliminate other issues first - in this case the WB - so that one can be sure, that the problem is not from that corner. So, don't be surprized if the banding remains there after the WB correction.

Btw, are these shots from the D90? (There is no Exif in the JPEGs.)

I've attached what I believe is a properly white balanced version (I had been trying to create a warmer/aged version previously, but the color banding issue made me stop and re-evaluate).

As you can see, the color issue remains, though slightly less offensive.

This shot is from a Panasonic Lumix-FX500 (P&S) - the previous image was from the D90.

[Panopeeper]

As you can see, the color issue remains, though slightly less offensive

Yes, I see it. How do those pinkish clouds look like in 100%? Can it be that there is strong noise there? What if you select the sky, only coarsly, and apply undusting to it and do not sharpen it?

I asked about the D90, because its raw data is lossy. Normally that does not cause any problem, but I have seen banding (with the D300), which was caused provably by the lossy compression. That was what I was thinking of when seeing the image you showed in the first post.

Another possible source of such banding can be color manipulation. Both the Selective Color Options of CS and the HSL adjustment of ACR caused me banding, because they work on selected colors, and the transition between the colors in the sky is very smooth, resulting in adjustments carried out on bands. It is difficult to explain, but try in this direction.

[Jonathan Ratzlaff]

Are these shot with grad filters?   Are they neutral?  The reason I ask is that there are areas that appear to be darker than one would expect; hillside on the far left and the ships masts.  Pretty heavy sharpening that  has affected the contrast as well.  This may be contributing to the issue.
The cloud image is a thunderhead.  There is often a colour cast at the base of the cloud.  Your post processing may have exaggerated it..

[TheSuede]

There's a known problem with some of the Nikon LR "Camera Standard" profiles that can give green/magenta shifting up through the brighter tones. Which camera/brand are you using?
Some more references can be found here:
http://forums.adobe.com/message/2405422#2405422

If Eric's around, he might give some more updated info on Adobe's standings. I gave up and started developing my own profiles. I could without doubt recommend doing this to anyone willing to spend an hour now on saving lot's of time later on.

[grammaticals--- :-)]

[SunderJ]

Are these shot with grad filters?   Are they neutral?  The reason I ask is that there are areas that appear to be darker than one would expect; hillside on the far left and the ships masts.  Pretty heavy sharpening that  has affected the contrast as well.  This may be contributing to the issue.
The cloud image is a thunderhead.  There is often a colour cast at the base of the cloud.  Your post processing may have exaggerated it..

All - appreciate the feedback

Jonathan: I did add digital grad filters in PP which produced the darkened left side of the hill and the darkened masts as you correctly pointed out.

I haven't done a detailed post processing yet - on both shots I noticed the color banding and wanted to first try to resolve that fundamental issue before spending more time.

Regarding the lossy nature of the D90 raw is an interesting point - both the Raw and jpegs from the D90 (on which the hillside was shot) exhibit this problem)

Suede - thanks for the link to the adobe forum - I had missed that in trying to research this problem - they are clearly discussing the exact same issue. Perhaps I should look into NX though pointing to raw conversion still doesn't explain how the problem could have creeped into the ship picture taken on a P&S (unless the in-camera raw conversion is to blame in that situation).

Pano: By undusting are you referring to removing noise / dust? Interestingly, Jeff had suggested adding grain, unevenly channel by channel to remove the banding exhibited in the picture even beyond the color issue. I'm still working on understanding/applying this solution.

[Schewe]

Interestingly, Jeff had suggested adding grain, unevenly channel by channel to remove the banding exhibited in the picture even beyond the color issue. I'm still working on understanding/applying this solution.

This is a grain making technique (it's actually in PhotoKit–more or less)...

Duplicate the background or make a merged visible copy and set it to Luminosity blend mode..

Select the Green channel and target it. Run the Noise filter...the amount will depend entirely on the file resolution but try 8-12
After running the Noise filter run Gaussian Blur and blur .3...

Target the Red channel and repeat but with less noise. If you used 8, try 6. Do the same G-blur

Target the Blue channel and this time increase the noise a bit. If the original green channel w 8, then try 10 or 12. Again do the blur...

Then look at the layer set in Luminance blend...vary the opacity to get the strength you want. You can also go into the Blend If options and taper off the blend in the extreme highlights and deep shadows...done proper, this noise can look very much like film grain and help break up posterization banding...