Hyundai Paint.

[Justinr]

I ventured out into the dull winters light to take some pictures of Hyundai diggers at a dealers today. The camera was the Nikon D3 with 24-85mm lens which I am finding well suited to the task. However, when I got back and viewed the images I noticed that the machines had all turned orange instead of a bright yellow. After much messing around I managed to retrieve some semblance of reality in Faststone by upping the highlights slider in the 'Adjust lighting' window to its full amount (I could not achieve anything satisfactory in Raw Therapee).

Has anyone any idea what might cause this? I rather suspect it might be the paint because I have noticed on other sites (including the dreaded FB) that a lot of images are suffering the same problem. Interestingly it doesn't happen with New Holland machines under similar conditions.  I attach some examples below -

No 1 is straight from the camera

N0 2 is adjusted in Faststone

No 3 is straight from the camera as well, but this time the yellow is much closer to reality. However, note how the boom in the distance has turned orange in the image when it's all the same colour in life. The camera settings were identical to the others.

[D Fosse]

2 and 3 are totally blown out, which clips the red channel so much that the green channel "catches up". Thus it becomes yellow. The boom that remains orange is darker, so it doesn't clip as much, and therefore doesn't turn yellow to the same degree. IOW these effects are all artifacts.

At the risk of stating the obvious: The underlying issue is that every image is processed somewhere, whether automatically in-camera or by the photographer in a raw converter. There is no such thing as "accurate" color rendition out of a camera (though surprisingly many people believe that). So there's no reason to accept the automatic processing in the camera, or the default rendering in a raw converter. That's what the sliders are for. If color and tone is critical you can use a target such as the ColorChecker Passport, but even then some tweaking will likely be necessary.

[Justinr]

I really don't think it's as simple as that. Increasing the exposure in either PS or Raw Therapee just makes the orange lighter, it doesn't restore the yellow, nor does adjusting the colour temp . I have other images where the yellow and orange appear on the same machine in the same light and as I mentioned earlier the problem does not occur with New Holland machines which are a similar bright yellow. I'm thinking there may be a component in the paint that the camera sees differently to us. BTW, I had the camera set to +1/3 exposure compensation.

[Slobodan Blagojevic]

I think that certain sensors react funny to certain colors (i.e., can not handle them well). I recently had a similar experience with a skirt. Purple in reality, it turned completely blue in out-of-camera (other colors in the scene were true-to-life). The only solution was to move Purple Hue slider all the way to the right (+100) to achieve the original purple.

[Justinr]

I think that certain sensors react funny to certain colors (i.e., can not handle them well). I recently had a similar experience with a skirt. Purple in reality, it turned completely blue in out-of-camera (other colors in the scene were true-to-life). The only solution was to move Purple Hue slider all the way to the right (+100) to achieve the original purple.

My only consolation is that I'm not alone, a quick search for Hyundai excavator images will show  more orange than yellow machines.

[fdisilvestro]

The color of the digger is likely outside of gamut even for Adobe RGB(it's the same issue with school-bus yellow). In this case the color is mapped to an inside-gamut color, which may look close or not to the original. One solution is to play with the individual HSL sliders for each color until you get the desired color.
It is very likely that the most similar color will have a blown out red channel

Regards

[D Fosse]

I'm thinking there may be a component in the paint that the camera sees differently to us.

Sensors have infra-red and ultra-violet sensitivity in addition to visible light. Manufacturers try to limit that sensitivity, but all filters produce a gentle roll-off so it can't be cut abruptly at a certain wavelength. Many pigments reflect very strongly in IR/UV.

So that's probably the direct explanation, but my point was that this happens often enough that, as I said, there's no particular reason to accept the out-of-camera rendition.

Here's one "classical" example of IR-sensitivity, OOC top and correct bottom:

[Justinr]

Sensors have infra-red and ultra-violet sensitivity in addition to visible light. Manufacturers try to limit that sensitivity, but all filters produce a gentle roll-off so it can't be cut abruptly at a certain wavelength. Many pigments reflect very strongly in IR/UV.

So that's probably the direct explanation, but my point was that this happens often enough that, as I said, there's no particular reason to accept the out-of-camera rendition.

Here's one "classical" example of IR-sensitivity, OOC top and correct bottom:

How did you correct the flower?  I've found the next best thing to a cure in my case is PS -> Selective colours -> Yellow channel -> take the magenta slide over to the left by at least 60 and then increase the lighting with curves, adjust saturation and contrast to taste. The image needs to be slightly over exposed in the first place for best results. It turned out to be a 30 second fix after spending hours messing around in RT and Fastone! Hey ho, all good fun I guess.

[D Fosse]

Actually, to be honest, hue shifts are a bit tricky to correct. Not because there isn't an abundance of tools to deal with it, but because most of them have a tendency to break up image integrity by shifting one pixel, but not its neighbor. Either/or. The result is often ugly banding and unnaturally sharp transitions, as well as exaggerated noise and grain. Hue/saturation falls in this category, both in Photoshop (even if you extend the range fall-off considerably) and ACR/Lightroom. I try to avoid all of these if possible.

My preferred tool for this is Selective Color in Photoshop. It works differently and looks for color components rather than absolute colors, so it's not nearly as destructive. In this particular case I removed cyan from blue. Magenta is fine here, it's the blue component (which represents the IR contamination) that needs to be killed.

Selective Color usually works very well for this, and it's flexible in that you can either "push" or "pull" a particular component. And once you get used to how it works, it's fast and intuitive.

[TMARK]

FYI, I bet a licensee of either Devoe Coatings or ICI Industrial Paint makes that paint at a facility in South Korea.  I shot collateral for Hyundai's ship building arm in 2000 and 2001.  That is what I remember about their paint.  You could prabably find a spectral reference for the paint from devoe or ICI.  Devoe provided me that information for paint they used on the interior of a vessel's balast tanks that would not reproduce on Fuji Provia, I believe it was. 

[Justinr]

FYI, I bet a licensee of either Devoe Coatings or ICI Industrial Paint makes that paint at a facility in South Korea.  I shot collateral for Hyundai's ship building arm in 2000 and 2001.  That is what I remember about their paint.  You could prabably find a spectral reference for the paint from devoe or ICI.  Devoe provided me that information for paint they used on the interior of a vessel's balast tanks that would not reproduce on Fuji Provia, I believe it was. 

I can imagine there being all sorts of problems with colour rendition in the dark bowels of a ship where you are relying purely on artificial light. From my experience with the diggers the colour seems highly dependent on light intensity with a 1/3 of a stop making quite some difference, it appears to be more a threshold than a gradual transition in this case but without going back and doing measured tests it's difficult to say for sure.