5

I was at the NEC today dropping off a car for an auction, and took this quick photo (no intention to use for anything so no attempt at composition!)

This was taken on a LG Nexus 5 mobile phone.

I have noticed this before on this phone, bright (not metallic) red, in direct sunlight seems to come out orange! I have also seen this effect on a red painted engine block.

How can this be explained technically?

This does NOT happen on my D800 or Hasselblad H4D, so must be related to this sensor(?)

Edited to add: the light here was through a window, so is unlikely to be UV related.

enter image description here

13

How can this be explained technically?

Auto Exposure and Auto White Balance.

The camera is trying to expose the image properly, but there's a huge difference in brightness between the shaded areas (most of the scene) and the foreground that's lit by strong direct sun. In order to get most of the image exposed correctly, it has to overexpose the car door and fender.

In addition to two different light intensities, the camera also has to deal with two different color temperatures. Most of the scene is in shade, where the ambient light has a strong blue component, while the bright sunlit areas have much more red and yellow. In order adjust the white balance to suite most of the photo (so that the Thrifty truck appears white instead of blue, for example), the camera reduces the blue across the entire image. This warms up the already-warmer-than-the-rest-of-the-image car door, and if the reds and yellows in that area weren't already blown out they will be after the white balance is adjusted.

Essentially, what you're seeing here is the same effect that photographers use in reverse to get deep blue backgrounds. The technique works like this: working outdoors, put a warming filter (like a "color temperature orange" gel) over a speedlite that's used to illuminate your subject. Set the camera's white balance to the tungsten (3200K) setting, so that the camera assumes incandescent light. That shifts the whole image toward blue, giving your subject normal looking skin, but making the sky much more blue. In the case with the cars, it's working the other way: you've got very yellow/red light in the foreground, but the white balance is set to compensate for blue, so you get even more yellow/orange/red in the foreground and normal-looking background.

  • Initially it seemed to me like there was more going on than just this, because of the weird gradients, but on second look, it could just be the fuzzy edges of the shadows being mangled by noise(+denoise+JPEG). Nice elaboration of what's unquestionably taking place, either way. – junkyardsparkle Mar 28 '15 at 6:01
13

It's because the red channel has completely blown out, whereas the green channel hasn't. (Nor blue but that's not having any effect, here.)

Suppose the true colour of the car is five parts red to one part green (the box on the left, below). If you underexpose the photo, you might find that the red channel's running at 50% intensity and the green channel would be correspondingly at 10%. Expose an extra stop and that goes to 100% red and 20% green, still showing the correct colour. But now if you expose another stop on top of that, the green doubles again to 40% intensity but the red can't double to 200% – it's stuck at 100%. So you end up with 100% red versus 40% geen, which is a ratio of five parts red to two parts green (the box on the right): much more orange than the true colour.

enter image description here

You see the same effect in photographs of the sky on bright days: the blue channel gets to 100% before the green channel, so the sky turns more cyan than it should be.

2

Looking at an RGB histogram of the image, I'd say the red channel is blown.

enter image description here

Hard to get deep reds in bright light, even when the overall exposure is a mid-tone, the reds blow out. I've had many pictures of dark red roses for example that look pink or orange.

  • Yes, the red channel is absolutely blown, but it's odd that parts of the blown area have red values <255, although I suppose this could be attributed to noise and compression artifacts... incidentally, there's an interesting answer related to red flowers and their infrared reflectance characteristics here. Not sure how much it applies to other red objects, though. – junkyardsparkle Mar 28 '15 at 5:41
  • @junkyardsparkle: I'd guess that the near-red values are almost certainly due to JPEG artifacting. One factor which makes JPEG uncompression and recompression lossy is that compression may cause the "mathematical" values of pixels which were both 255 to become 253 and 257 (note the average value is still 255), but uncompression may turn the 257 to 255 (reducing the average to 254). – supercat Mar 28 '15 at 18:08
1

My guess is that the phone's JPEG engine is trying to recover blown-out highlights in a way that results in this coloration artifact (as opposed to simply clipping them to white). It's probably trying to work with the color information it has, which would be the red around the overexposed areas, and possibly the yellow cast that direct sunlight would have after white balancing the image for the shaded part. Possibly using a method similar to what's described here.

-1

I have been researching this phenomena of red paint changing in the sun. IMO it may have nothing to do with the camera, but everything to do with the color itself. In searching for a red for my 55 Chevy restomod I examined many shades of red in paint stores, at the shop that is painting my car and everyone changes to orange or some similar color in the sun. Looking at them indoors, in the shade, later in the evening, they all change color from the deep deep red that I was after.

I went through this with my last project car and the same thing. Looked like a very dark deep red in shade but in light looked like a race-red with more orange. I am still researching this as it is driving me nuts, but I do know if you are looking at painting a car pick the color between 10AM and 2PM. Not durning dusk or shade etc. Hope this helps some.

  • No, it's not the paint. Just the classic effects of blowing out a channel. As an example you might have a nice red at RGB(210, 50, 35) but the portions illuminated by sunlight might have say 10 times the Lux. Assuming the color temp of "white" is the same, that would increase each of the RGB values by about 4 (reduced by gamma) to an RGB of (840,200,140). Since the channel maxes at 255 you wind up with RGB(255,200,140) which is a fairly bright orange/yellow. – doug Apr 16 '18 at 5:13
  • 1
    Hi TerryBerg, Welcome to PSE. Yes, an object can only reflect what wavelength shown upon it. You are viewing the colour chips under different conditions of varying wavelengths. They should appear differently. In addition is the actual composition of the pigments—the term is metamerism. Your powers of observations are better than most that I've met. BTW, paint samples are the most challenging of all problems in colour management. The highest degree of instrumentation is necessary when working with this medium. To make matters worse, everything changes after the finish is baked to harden it. – Stan Apr 16 '18 at 19:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.