Sunsets are difficult to photograph, and green flashes are even more difficult to capture. The brightnesses of the Sun, the sky, and the foreground usually differ by many orders of magnitude, and differ by additional large factors from one sunset to another. Even within a few minutes, the brightness can change drastically. At moderate latitudes, the correct exposure changes by about a factor of 2 (one full stop) every minute.
In addition to the difficulty of getting the exposure right, there are several deceptive effects that can lead an unwary observer to suppose that a green flash has been captured, when in fact all the picture shows is some quirk of the imaging method used. These misleading appearances in pictures are generally called artifacts.
Regardless of the technology used to capture the image, there are inherent imperfections in the optical image itself that occasionally are visible.
For example, Poul A. Costinsky has a photograph on the Web that shows a bright-green streak in the upper left corner. It isn't a green ray, but what optical workers call a “ghost” — an internal reflection in the lens itself. In this case, the broadband anti-reflection coatings used on his Canon 55 mm lens were designed to minimize reflections across the spectrum by reducing them to the lowest possible value in the red and blue; but this leaves a little reflection in the middle of the spectrum (i.e., in the green). This residual reflection is very weak, but it's visible in his picture because the Sun is so overexposed.
Ghosts like this are typically seen in the image at a position diametrically opposite to the overexposed Sun. In the picture mentioned above, the Sun is in the lower right corner, and the ghost in the upper left. This reflection about the center of the image is a dead giveaway that the spurious image is a ghost.
Such internal reflections are more common in zoom lenses, because their large number of elements (and surfaces) provide more opportunities for reflections to occur.
Mila Zinkova has a really splendid example of ghosts, in her picture of crepuscular rays at this web page.
Silver-halide photography has been around for all of our lives, so we're accustomed to its quirks. Experienced photographers are aware that an overexposed image of the low Sun will turn out yellow, or even white, on color film, even though its nominal color is reddish orange or red. This effect of overexposure is due to saturation of the film; because the Sun is reddish, the red-sensitive layer saturates first.
Another artifact of overexposure in color photography is that an overexposed green flash shows very washed-out, pale colors on reversal films like Kodachrome and Ektachrome. Underexposure is preferable, if you want to capture the highly saturated colors typical of green flashes. But even at best, the colors that can be reproduced on film are much less saturated than the colors of real green flashes.
Other photographic artifacts that might show up are halation and solarization, though the latter requires very extreme overexposure.
The point is, the photographic process has been around long enough that its vices are well known, and experienced photographers will be aware of them. If you know what can go wrong, and keep an eye out for it, you aren't likely to be fooled.
These days, digital cameras are almost as common as digital watches, the fad of a generation ago. But digital cameras are new; we're not used to their quirks. So it's easy to be deceived when the camera does something it was designed to do, but isn't want you want it to do.
A common example is automatic adjustment of color balance. Most digital cameras look at the relative amounts of the primary colors in the scene, and then try to adjust the color to make things look “right”. If you take a picture in open shade, where the illumination comes from the blue sky, the camera reduces the brightness of the blue sub-image and increases the gain for the red, so that objects in the picture come out with approximately their “correct” colors. If you take a picture indoors by incandescent light, the camera compensates for the reddish illumination by boosting the blue and reducing the red content of the picture.
Of course, the eye does all this compensation automatically; it's not so easy for the camera. That's why we had “daylight” and “tungsten” films in silver-halide photography, and professional photographers carried around a set of color-compensating filters to handle the variations in lighting. The fact that electronic cameras can compensate for the lighting automatically is often regarded as one of their big advantages.
However, when you try to photograph a sunset, the Sun is very red, as are any clouds low on the horizon. Often, a digital camera will look at all this red stuff in the picture and decide to shift colors in the opposite direction. Sometimes the effect of that is to produce a picture with a pronounced greenish cast.
Sometimes people unfamiliar with this quirk will discover a greenish picture in a series of images they've taken of a sunset, and jump to the conclusion that they've photographed a green flash — or, better yet, one of those rare “green ray” displays. Sometimes they send them to me and ask if this is really so, and I have to explain that, no, that's just the way the camera works. It can shift the balance from one picture to the next, and back again, depending on just how much of the red cloud is included, or where it is in the frame.
The camera maker would insist that this isn't a bug, it's a feature. Well, 99% of the time, for most people, it is a feature. But for those of us looking for green flashes, it's a bug.
A related problem is automatic exposure adjustment. Many digital cameras don't really adjust the exposure at all; they just adjust the gain of the image so that a tiny fraction of the pixels are saturated. Those saturated pixels might be on the Sun itself, or a cloud near it. Once again, the result is a funny-looking picture.
Often a big giveaway that something isn't really in the sunset sky is the fact that the photographer (and other people nearby) didn't see anything unusual at the time the picture was taken. Now, real green flashes are pretty spectacular: people notice them. It's hard enough to capture the ones you actually see. If you didn't see a flash, that's usually because there wasn't one. The camera is very unlikely to catch a flash missed by everyone watching the sunset (unless a very long telephoto lens was used).
Sometimes it's possible to re-adjust the color balance of a greenish picture in such a sequence to establish that the color is just an artifact of the color balance. But this is tricky to do, because digital images are usually stored in a way that displays correctly on a computer or TV monitor. But these monitors usually produce brightnesses that are roughly proportional to the square of the signal fed to them. So a compensating nonlinearity is built into the digitized image.
That means that we have to undo this nonlinearity to get back to numbers that are proportional to brightnesses, before we can play games with color balance by adjusting the gain of the R, G, or B channel. And then, you find that the overexposed red channel had its data truncated at 255, so you can't make a realistic adjustment of the truncated values. So it often isn't easy to process the image to show how the phony green color was produced.
A different kind of nonlinear problem is shown in a sunset image taken by surfer J. H. Fleming. Here, the overexposed area of sky near the Sun has driven the detector into nonlinearity as it approaches saturation, so it is no longer responding in proportion to the light level. Of course, the very red light of the low Sun makes the red-sensitive pixels become nonlinear before the green ones. In this picture, there's an annular region where the red pixels are near saturation, but are not yet “clipped” at digital values of 255. The automatic white-balance mechanism of the camera has decreased the gain of the red pixels, which accounts for the orange (rather than deep red) hue of the outer parts of the solar aureole. Closer to the Sun, the red pixels are nonlinear and are only slowly increasing their output, but the green ones are still responding fully to the increasing brightness. Because the gain shift has reduced the red, relative to the green, in the output image, there is a ring in which the rapidly increasing green values overtake the red, producing a pale-green color in the image. So the nonlinear red response interacts with the gain shift and the (still linear) green response in the next zone of the aureole closer to the Sun to produce an unrealistic greenish-hued ring. Finally, the image becomes white where all the pixels get clipped at values of 255, right next to the Sun.
Photographic film has even more of this nonlinearity as it approaches saturation; but it never produces this kind of hue shift, because films don't have the automatic white adjustment of digital cameras. So here again we see a new kind of artifact in digital cameras that the older chemical photography lacks.
Another artifact of digital cameras is due to charge spill in CCD detectors. Bob Collin, of Beaverton, Oregon, sent me this fine example:
What you see here is leakage along the columns of the CCD, caused by the
overexposed solar image. I'd have expected this streak to appear red, but
maybe the green hue is due to a color-balance shift of the kind discussed
above: notice that the clouds around the Sun appear yellow, rather than
the reddish orange you'd expect.
In any case, this is a camera artifact, not an atmospheric phenomenon.
What's happening here is that the Sun's bright image produces vastly more photoelectrons than the maximum capacity of the little “electron wells” in the chip that hold and transport the charges forming the image. It's sort of like those plastic ice-cube trays that have little grooves between the compartments, so that you can run water into one, and it will progressively flood the others. Here, it's electrons instead of water, but the overflowing process is analogous: excess charges flood the column, producing a bright artifact in the image, as if it had been exposed to a vertical strip of light in the image plane.
Just to show that this isn't an uncommon problem, here's another image
with a green charge-spill artifact, provided by Lou Adzima here in San
Diego County. This image was “taken 8/30/06 with a D70 Nikon camera
with a 18-200MM lens with a normal UV filter”; a frame taken 4
seconds later, but with the Sun out of the frame, shows no such green
stripe.
[The black stripe at the bottom is a different side-effect of extreme overexposure. Apparently the intense illumination at the solar image has made the column “leaky”, so that it failed to shift out all the image charge from the more distant rows, which have to be passed along the electronic “bucket brigade” more times than those at the top. So electrons that should have contributed to the brightness at the bottom of the affected columns failed to appear in the readout, and this area came out black!]
Yet another kind of digital-camera artifact is due to excess infrared response. Les Cowley shows a nice example on this page: a camera has produced an infrared image of the low Sun when the extinction was so great that it was invisible to the eye. Of course, the purple color is entirely artificial.
Some artifacts of digital photography are due to the software used. Jason Carter has found an egregious example that's obviously an artifact, not a green flash. It turned out that this was a glitch in the software used to process the image: Jason says, “I have confirmed that the picture defect is caused by RAW processing in Apple's Aperture version 1.1. Processing the images with version 1.0 does not generate any defects.”
So common are the artifacts of digital photographs that one Web page lists dozens of pieces of software that are available to “correct” the commoner ones. A quick search with Google turned up additional pages that discuss these problems; most are devoted to the rather obvious aliasing and compression artifacts, but others are mentioned. There's also a comparison of digital and film photography.
Things are even trickier with videos, because video images go through much more complicated processing. Some of this involves lossy data compression, which throws away some of the details of the actual scene. But in addition to that, the data are separated into brightness and color (for reasons historically connected with making color TV compatible with the black-and-white system that preceded it); and the color information (called “chroma” in the video world) is separated into two opponent channels, red-green and blue-yellow, even though the image capture is usually done in terms of RGB.
This opens up further possibilities for mischief, if anything goes wrong in the electronics or the recording apparatus. And of course most video cameras also have the automatic color adjustment (known as “white balance” to the video people) that digital cameras have. So there are more ways for things to go wrong.
A nice example of this was sent to me by Alan Dean Foster, who took a sunset video in Australia. His sunset sequence shows occasional frames in which the red sky at the horizon turns bright green. At first glance, this looks like a real phenomenon, because the sky and sea are still nearly the same color in the green-horizon image as in adjacent red-horizon ones. Here are his examples:
and
I tried adjusting the color balance, and quickly found that no possible shift in color balance could turn one frame into anything like the other. So the green image is clearly not due to a color-balance artifact.
However, inspection of the video frame by frame shows that the green frames appear suddenly and very briefly, with no smooth transition between red and green frames. Even the briefest green flash lasts several tenths of a second; but these green frames appear from nowhere and disappear just as abruptly; some are just single frames sandwiched between normal red ones, and the longest green sequence lasts only a tenth of a second. This certainly isn't anything related to green flashes.
Furthermore, each green frame has some garbage at the bottom edge, indicating some synchronization problem in the video system. (You can see this at the bottom edge of the green frame above.) And it's preceded by red frames with the same problem. This strongly suggests a video artifact of some kind.
It occurred to me that maybe the sense of the red-green chroma signal was getting reversed, so that what should have been red was displayed as green and vice versa. If that were the case, I should be able to re-create the green frame, or something like it, simply by disassembling the red frame into its red, green and blue components, swapping the red and green, and then re-combining them into a color image. (On Linux, it's easy to do this with the ppmtorgb3 and rgb3toppm commands.) I did this; here's the result:
and
On the left, you have the red frame with the red and green swapped; on the right, the original green frame from the video. I think the similarity is close enough to establish this as the correct explanation.
The reason the sky (and sea) looked nearly the same in both red and green frames is that the sky is mostly blue, with very little red or green content. So swapping red and green hardly changes the sky at all.
The lesson to be learned is: don't trust digital or video images, unless you know all the ways they can go wrong.
Copyright © 2004 – 2008 Andrew T. Young
Back to the ... photographic advice page