Help with magenta/green crossover

Hi,

I've never had this happen. This is a screenshot of a white painting on a white wall from Photoshop at 200 percent. A little mixed lighting, but had this on a strobe-only (white umbrellas) image too. Nikon D610, Legacy 55mm lens I've used forever, SanDisk card, processed with ACR in CS6, Mac

Any ideas of what happened? Color issues on painting and wall.

Any ideas how to fix it? Cannot reshoot. I've spent all day with magenta and green adjustment layers, going nearly pixel by pixel

Thanks

You have a raw?

Yes. The coloration is visible in the dng when I open it in ACR. My colorchecker passport profile was also odd, and didn’t seem to make any difference

Convert to B&W?

Good idea, but parts have subtle color that needs to be included

How about posting a full rez photo in "Digital Darkroom" "Post Processing Challenge" and letting those guys have a crack at it.

Good idea but I’m away for the weekend and don’t have access to my images. I’m going to try messing with white balance in ACR, and working on them as 16 bit, which I should have probably done. Wish i could figure out what caused this

What exactly are we looking at?

The image posted shows faint mauve squares at full-rez, which looks like posterisation from the contrast boost and JPEG compression. But I don't see any drastic green to magenta shift with brightness.

I suspect your monitor might have gone out of calibration. Have you tried viewing the image on another monitor?

What does the eyedropper tool tell you about the RGB ratios?

How about posting a full rez photo in "Digital Darkroom" "Post Processing Challenge" and letting those guys have a crack at it.

Better yet, the raw. That's why I asked if it existed instead of a rendered image.

What exactly are we looking at?

 

The image posted shows faint mauve squares at full-rez, which looks like posterisation from the contrast boost and JPEG compression. But I don't see any drastic green to magenta shift with brightness.

 

I suspect your monitor might have gone out of calibration. Have you tried viewing the image on another monitor?

 

What does the eyedropper tool tell you about the RGB ratios?

It does look a little like posterization ...

I am away from my computer for the weekend, will check the ratios on Monday. Should I check the mauve areas vs the green? I do not know as much about color management as I should.

I have looked on my other monitor, that was one of my first thoughts. The odd magenta and greenish areas are still there.

What I posted was a screenshot png, from a 300 percent enlargement of the dng in camera raw. It’s like the base is green, and any image detail is that mauve color.

When I re-did the white balance after my original post, it got better but still there.

Thanks so much.

Ahh, a screenshot.

Well, I'm afraid that's not going to tell us much.

The mauve squares could be an artefact of a bad video card LUT or incorrect colour-profile or practically anything else in the processing chain between camera sensor and screen.

The thing is that the mauve pixels don't appear to vary much with their base colour, which makes me think it's an artefact added in the computer.

If it was, say, a colour cast from a lighting imbalance or shadow contamination, then I'd expect the cast to be stronger in the shadow or highlight region. So it looks like something is being 'added' in the post-processing chain.

What and where exactly is pretty impossible to guess from a screenshot. We'd need to see the actual camera image.

An afterthought: Any chance that the highlights were blown in the green channel? That'd probably do it.

Ahh, a screenshot.

Well, I'm afraid that's not going to tell us much.

 

The mauve squares could be an artefact of a bad video card LUT or incorrect colour-profile or practically anything else in the processing chain between camera sensor and screen.

 

The thing is that the mauve pixels don't appear to vary much with their base colour, which makes me think it's an artefact added in the computer.

 

If it was, say, a colour cast from a lighting imbalance or shadow contamination, then I'd expect the cast to be stronger in the shadow or highlight region. So it looks like something is being 'added' in the post-processing chain.

 

What and where exactly is pretty impossible to guess from a screenshot. We'd need to see the actual camera image.

 

An afterthought: Any chance that the highlights were blown in the green channel? That'd probably do it.

Thanks. This makes me feel a bit less worried. Though there may be a lighting issue, since there are some magenta streaks around where my umbrellas were hottest.

Re: your afterthought, I think exposure is involved but it’s confusing. The most affected areas are pure white with almost no texture, but they’ve been underexposed in camera by about 2 stops, because other areas are a very reflective material that I didn’t want to blast completely out. Fun.

I will post a better image on Monday. Thank you so much!

 

Re: your afterthought, I think exposure is involved but it’s confusing.!

Without the raw, we cannot tell you anything about exposure. That would require examining the raw's actual Histogram in a product like RawDigger that actually provides a raw Histogram. We could examine each channel and know without question about clipping of the actual raw data.

RawDigger

Again, upload a raw, some here can tell you exactly about both the exposure and neutrality of the capture.

Hi. Sorry, I can't figure out how to post the dng file...I can't really put it on my website....

Sorry, wasn't thinking.

Here it is:

Dropbox - PG_19__DSC3052.dng

Please don't share this image off this forum. Thanks.

First off, the image isn't clipped in any color channels based on viewing the raw Histogram and image in RawDigger. Depending on your goals (all I see is a wall and some color targets), one could suggest it's a quite a bit under exposed; you can 'expose more to the right' but that's a different story. According to RawDigger, even 1 additional stop of exposure at capture, not close to clipping any channels.

Opening in ACR, there is a missing camera profile (MC_Strobe) so I cannot view as you did. So I pick Adobe Neutral. Zoom in and white balance on the white patch of the Kodak target. The readout's of the wall are then approx. 170/170/172 which is pretty neutral. I see no banding at 100%.

The image a rather noisy, mostly due I suspect to the far from ideal exposure for raw data.

First off, the image isn't clipped in any color channels based on viewing the raw Histogram and image in RawDigger. Depending on your goals (all I see is a wall and some color targets), one could suggest it's a quite a bit under exposed; you can 'expose more to the right' but that's a different story. According to RawDigger, even 1 additional stop of exposure at capture, not close to clipping any channels.

Opening in ACR, there is a missing camera profile (MC_Strobe) so I cannot view as you did. So I pick Adobe Neutral. Zoom in and white balance on the white patch of the Kodak target. The readout's of the wall are then approx. 170/170/172 which is pretty neutral. I see no banding at 100%.

The image a rather noisy, mostly due I suspect to the far from ideal exposure for raw data.

Thank you so much for taking the time to look at this.

The area in the center is coated with a fluorescent material that only reacts with a pretty direct light, so I planned to adjust the flat white side panels digitally.

Thanks for the reassurance. I will get my screens recalibrated and keep going.

I appreciate your help.

You first major issue is a good 1.5 to two stops under exposure for this capture. Look at the RGB values of just the white patch of your Kodak target, and the Histogram of the entire capture:

http://digitaldog.net/files/UnderExposed.jpg

Hi Paula

I am not sure exactly what I am looking at with your image yet the Hue map below may be of some help as a clue to the source of your problem (so in other words I am guessing at your set up and what I infer from this Hue map. I am assuming you made the shot from an angle that gives a pretty significant angle perspective. If this is not the case you can ignore my inputs below.

All of the Hue variations are very subtle which are highly magnified with this Hue map.

Not that the middle band of Hue variation does not follow the same perspective as the edges of the white painting/wall. This suggests to me that the problem lies not in your lighting yet something inside your camera. Not sure if this has to do with some aberrant lighting that made it into the camera or something to do with the shutter or sync settings yet having a vertical band when I would expect a matching perspective looking band points me to something with the camera or camera settings. Though possible yet less likely it is some indirect light aberration entering the camera yet given the area of question is illuminated equally with the plane of the sensor and the edge lines exactly lined up with the shutter edges I think that less likely.

Hope this gives you a direction to consider yet I am reaching a bit given my lack of detailed understanding of your set up.

Dropbox link follows of Hue Map image (reduced to 1500 x 1000 pixels)

Dropbox - Off-White-Image.jpg

ADDED: Just saw you post about the center area being coated differently. That certainly could explain the Hue band if it matches that area of the image in the link I posted

Hi Paula

I am not sure exactly what I am looking at with your image yet the Hue map below may be of some help as a clue to the source of your problem (so in other words I am guessing at your set up and what I infer from this Hue map. I am assuming you made the shot from an angle that gives a pretty significant angle perspective. If this is not the case you can ignore my inputs below.

 

All of the Hue variations are very subtle which are highly magnified with this Hue map.

 

Not that the middle band of Hue variation does not follow the same perspective as the edges of the white painting/wall. This suggests to me that the problem lies not in your lighting yet something inside your camera. Not sure if this has to do with some aberrant lighting that made it into the camera or something to do with the shutter or sync settings yet having a vertical band when I would expect a matching perspective looking band points me to something with the camera or camera settings. Though possible yet less likely it is some indirect light aberration entering the camera yet given the area of question is illuminated equally with the plane of the sensor and the edge lines exactly lined up with the shutter edges I think that less likely.

Hope this gives you a direction to consider yet I am reaching a bit given my lack of detailed understanding of your set up.

 

Dropbox link follows of Hue Map image (reduced to 1500 x 1000 pixels)

 

Dropbox - Off-White-Image.jpg

 

 

ADDED: Just saw you post about the center area being coated differently. That certainly could explain the Hue band if it matches that area of the image in the link I posted

Thanks, John. The painting is a trapezoid! And the band in the center (vertical stripe) is caused by a difference in the fluorescing material. I've never looked at a hue map; is this specifically for viewing color shifts? This was really hard to make it appear like it does to your eyes. That's why I chose to light it for the center and underexpose the rest. I guess the weird color is from grossly underexposing the rest of the scene in order to capture the fluorescent center. So glad I don't do this on film anymore...

You first major issue is a good 1.5 to two stops under exposure for this capture. Look at the RGB values of just the white patch of your Kodak target, and the Histogram of the entire capture:

http://digitaldog.net/files/UnderExposed.jpg

I am afraid I know little about sensitometry...the channel histograms are for the white patch? Is the red channel significantly shifted because my umbrellas skew red? I underexposed to maintain detail in the center fluorescing area. I think I would try to light it differently next time. So this is probably the source of my issue? You say "first major issue"....do you see another??

I am afraid I know little about sensitometry...the channel histograms are for the white patch? Is the red channel significantly shifted because my umbrellas skew red? I underexposed to maintain detail in the center fluorescing area. I think I would try to light it differently next time. So this is probably the source of my issue? You say "first major issue"....do you see another??

This isn't 'sensitometry'. The white patch is pretty massively under exposed. That's one reason why you see so much noise in your capture. You can't 'lighten' it and get anywhere. You need to better expose the capture IF your goal is the least noise and best data. It looks like you're exposing raw data as if it's JPEG data and that's a mistake and the reason why your capture is so under exposed and hence noisy. See:

http://digitaldog.net/files/ExposeForRaw.pdf

This isn't 'sensitometry'. The white patch is pretty massively under exposed. That's one reason why you see so much noise in your capture. You can't 'lighten' it and get anywhere. You need to better expose the capture IF your goal is the least noise and best data. It looks like you're exposing raw data as if it's JPEG data and that's a mistake and the reason why your capture is so under exposed and hence noisy. See:

http://digitaldog.net/files/ExposeForRaw.pdf

This article is very helpful, thanks.

Though I'm not sure I could have opened up much without blasting out the fluorescent portion.

Your brightest highlight in the capture is a good stop or more from clipping.

You should reshoot, bracket 1, 1.5, 2 and 2.5 stops more, then examine the raw in a demo of RawDigger. Then you'll know exactly how to target the exposure.