You’ll need red-cyan glasses to experience the 3D effects in this post. Don’t let that deter you if you don’t have any. I just ordered ten pairs for $4.89 (shipping included) on eBay and got them in three days. I’d like to think that’s a small price to pay and a short time to wait for the historical media novelties I’d like to share with you this time around. Granted, it’s not free, and your gratification won’t be immediate—that’s why I’ve been hesitant to rely on anaglyphs at Griffonage-Dot-Com in the past, and it’s one of the reasons why I’ve been experimenting so much with pseudo-3D animations that don’t require special equipment for viewing. But anaglyphs give a true 3D effect that the animations don’t, and they also offer a few other interesting strategic opportunities. Besides, it’s not like you can’t use red-cyan glasses to view other cool stuff online. So if you’ve got a pair already, grab them and read on; but if you don’t, order yourself some and bookmark this page so you can revisit it once they arrive.
3D anaglyphs consist of two superimposed images with different color filtering that corresponds to differently-colored lenses in a pair of glasses designed to view them. They’re ordinarily used to make stereoscopic illusions from pairs of images captured from horizontally separated vantage points. For example, here’s an anaglyph I made from a pair of tintype portraits taken using a pair of lenses mounted side by side (see here, here, and here for some other experiments with pairs of tintypes):
In this case, the two images probably weren’t intended by the photographer for stereo viewing—the goal would instead have been to make a few separate portraits at once—but apart from that, they’re no different in a technical sense from the image pairs published on commercial stereocards. Turning conventional stereoviews like these into anaglyphs is an attractive way of viewing the past in 3D, but it’s also a pretty standard technique—there’s nothing particularly novel or interesting about it. Still, we can do some other things with anaglyphs besides, and to push the envelope in the Griffonage-Dot-Com tradition, I’d like to illustrate a few more creative scenarios below.
For example, we can use anaglyphs to make 3D illusions from pairs of images taken from vertically separated vantage points as long as we rotate them ninety degrees, like this:
This anaglyph conveys just as striking an impression of spatial depth as the previous one, even if the image isn’t oriented “correctly” this time. I’ve cropped the two examples shown so far to omit the parts of the images that don’t overlap, but here’s another example with both tintypes fully shown:
The vantage points of pairs of “identical” tintypes seem to vary vertically as often as they do horizontally, so by adding sideways rotation to our anaglyph-making toolkit, we open up a lot more of this material for true 3D viewing—a prime example of what I’ve been calling eduction against the grain, i.e., taking inscriptions and making them sensorily accessible in meaningful ways their makers never intended. Of course, most pairs of tintypes were never “meant” to be seen in 3D. But when the vantage points of our tintypes vary vertically rather than horizontally, we can be even more sure than usual that we’re experiencing 3D effects snatched from the past without its conscious collusion. After all, nobody who had stereoscopy in mind would ever have chosen to set the images up to be viewed sideways!
Another trick we can perpetrate with sideways-rotated anaglyphs involves creating stereoscopic illusions from pictures of water reflections, as I showed in a previous post. To do this, we rotate the source image 180 degrees, flip it horizontally, overlay it on the unaltered original, and then rotate the combination ninety degrees, assigning the version with the reflection on the right to the left eye and the version with the reflection on the left to the right eye. I just acquired a fine new candidate image for this technique: a round photograph taken with an early Kodak. It’s undated, but the eBay seller I bought it from also had a number of others appearing to be from the same source with handwritten dates in 1889 and 1890, so I assume mine is from that period too, making it one of the first amateur snapshots ever taken with a simple consumer camera. Here’s an unaltered scan of the original print, which is a beautiful thing in its own right:
Is this the oldest example of a 3D effect recovered from a snapshot taken in the spirit of Kodak’s famous slogan, “you push the button, we do the rest”? I think it might be—but you be the judge.
Meanwhile, anaglyphs also give us a handy method for testing whether any two given images were created from separate vantage points, and if so, how they differed: vertically, horizontally, or both ways at once.
In May 2014, I blogged here about the perspectival relationship between two versions of the Mona Lisa: the one at the Louvre in Paris and the one at the Prado in Madrid. Claus-Christian Carbon and Vera Hesslinger had then recently put forward a hypothesis that these two paintings taken together form a stereoscopic pair, and maybe even an intentional one, since Leonardo da Vinci wrote about the phenomenon of stereo vision, and hence was plainly aware of it to some extent. Critics—e.g., the London Stereoscopic Company as expressed here on April 22, 2014, and Martin Arguin as quoted here—had already pointed out that the vantage points of the two images seem to vary vertically more than they do horizontally as needed for a conventional stereoview. I reached the same conclusion myself after rescaling the two images to keep the size of the background scenery consistent. But then I took things a step further by illustrating the apparent difference in vertical perspective using animated GIFs.
She leaps out at you pretty strikingly this way, doesn’t she? Carbon and Hasslinger present a couple of right-way-up anaglyphs in their formal paper, but these are limited to the face and hands:
They don’t provide any anaglyph of the complete scene (although they do offer us an animation), maybe because there isn’t any way to line up the whole images to produce a decent 3D effect without rotating them sideways, as I’ve done. Of course, there’s nothing mutually exclusive about vertical and horizontal displacement, and an anaglyph of an image pair can yield a 3D effect oriented both vertically and horizontally if the vantage points varied along both axes. Here’s an anaglyph of a pair of tintypes that does a nice job of illustrating the point, first shown rotated sideways, and then shown right way up for comparison:
In this case, the two lenses were displaced vertically more than they were horizontally, so the sideways anaglyph seems to have a little more spatial depth than the right-way-up one. If we extend the same principle to the two Mona Lisas, we could say that the discovery that they “work” together so well as a sideways-rotated anaglyph doesn’t necessarily mean there wasn’t also enough horizontal displacement in their vantage points to produce localized right-way-up stereoscopic effects in areas such as the face and hands that received especially meticulous attention. It’s worth bearing in mind here that paintings aren’t photographs; imagine the results we might expect from freehand copies of the tintype pair we just examined. Maybe the horizontal difference in vantage points between the two Mona Lisas was great enough to sustain a perspectival illusion spanning the whole scene, even considering the vagaries of subjective manual depiction, whereas the vertical difference wasn’t. In any case, I think the question here should be not whether Leonardo intended the two Mona Lisas to function together as a stereoview—it’s clear he didn’t—but whether we can create cool perspectival illusions from them anyway, and what we can learn in doing so about how the paintings came into being.
We can create neat 3D illusions from even older historical source material using face-averaging software (see here, here, and here for some of my other face-averaging experiments). The software I’ve been using for this purpose, Abrosoft’s FaceMixer, lets me decide how much to weight the “feature” of each face separately from how much I weight the “shape” of each face. Here’s an anaglyph I made by averaging a large number of paintings of women’s faces dating between about 1440 and 1460, with the “right” image combining all the faces at their original angles (flipped vertically as needed), and the “left” image made by weighting the shapes of the most forward-facing images more heavily. Behold the average mid-fifteenth-century European painting of a female face in glorious 3D!
Of course, we can apply the same technique to more recent image sets as well. Here’s a preview of a project I’ve been working on for a few months now, in which I’ve been using face-averaging software to illustrate the evolution of the “fashionable face” over time. This particular example shows the average of 203 separate images published in Vogue during the first half of 1964:
My earlier strategy for displaying stereoviews online hinged on turning them into animations, substituting a time base for stereopsis. Thus, instead of presenting still images like this—
Animations have definite advantages: you don’t need any special equipment to view them, and they can do a better job of handling parts of the image (like the alligator’s tail) that diverge too much for the eye to resolve them comfortably in 3D. But they also have drawbacks. Fading back and forth between images is simple to do, but the most spatially divergent parts of the image pair will often appear to teleport from spot to spot rather than rotating smoothly in space. Morphing can eliminate that problem much of the time, but it requires time-consuming manual work (at least, using the techniques I’ve figured out so far), and some image pairs just don’t seem to work with it, as when foreground objects need to move conspicuously past background objects.
And then there’s one other drawback. Let’s say we opted to display our stereoscopic average of mid-fifteenth-century European paintings of female faces as a morphed animation rather than as an anaglyph, like this:
But what if, in addition to the spatial-depth effect, we also wanted to display a sequence of averages to illustrate how they changed over time, like this (covering the century leading up to the average shown above)?
We’d run into a dilemma because we’d be trying to connect the time base of the animation to two different parameters at once: passage in time and change in perspective. On the other hand, if we were to present each average in turn as a 3D anaglyph, there wouldn’t be any dilemma, since we’d free up the time base to represent time. I haven’t done that yet with any of my face-averaging sequences, but I do have a few other experiments to share below which combine 3D anaglyphs with animations.
Stereoviews were often produced in coherent series to tell stories, so I thought it would be nice to display one of these series over time as an animated GIF. The most suitable example of a stereoview series in my own collection is “A Hold-Up,” consisting of fifteen numbered but uncaptioned cards produced by the Mills Novelty Company (catalog number 101). A calendar visible on the wall in each scene is dated January 1906, which was probably when the sequence was photographed; and each card is also blind-stamped “APP. CHICAGO NOV. 9. 08 F.A.R.,” which I suspect had something to do with the cards being approved for display in coin-operated machines in late 1908. I suppose we could animate each scene individually and then string the animations together into a longer sequence, but anaglyphs let us show the program efficiently and effectively as a fifteen-frame GIF (plus one frame for the title and a black frame to mark the end). The rotational alignment of the images is a little wonky, but it reflects how they’re actually printed on the original cards—I didn’t take the time to “fix” anything, and simply chose the right edges of the cards and the back left corner of the table as my fixed anchor points.
In addition to intentional series such as “A Hold-Up,” we can sometimes find two versions of the “same” stereoview taken from slightly different vantage points, and with slightly different poses—the idea being that the more original negatives there were, the faster duplicates could be printed for sale. Here’s an animated anaglyph that alternates between two variants of a stereoview called “Rebels Defeated”:
We can also incorporate anaglyphs into more rapid animations of historical images, approaching a bit closer in spirit to 3D cinema. Here’s the final row of images on Plate 521 of Eadweard Muybridge’s Animal Locomotion, published in 1887. Unlike Muybridge’s better-known sequences that show motion over time, these six images were taken simultaneously from a set of lenses arranged in a semicircle.
Most of the adjacent image pairs in the sequence differ too much in perspective to create worthwhile stereoscopic illusions by themselves, but when the anaglyphs are all shown in rapid succession, they give a fair sense of a rotating three-dimensional figure. And we can perform the same trick with the well-known “autoportrait tournant” by Nadar (Gaspard-Félix Tournachon)—an “ordinary” animation of the images is shown on the left, while an animation based on anaglyphs of adjacent images is shown on the right:
The moon phase animations I blogged about back in March are also attractive candidates for this approach: in theory, we should be able to make the portion of the moon illuminated by the sun appear to rotate around it in 3D. Here’s my original close-up animation of moon phases as shown in the Catalan Atlas of 1375:
I’ll admit that I’m not sure how much of a 3D effect we get in either case, but staggering the views presented to the left and right eyes by one frame does seem to make the animations feel a little smoother, at least, maybe because single-frame disruptions such as the page split in the Catalan Atlas never have our binocular vision wholly to themselves. That’s something.
And whatever you make of these last couple experiments, I hope it’s still clear that we can harness anaglyphs in a number of unconventional ways to create stereoscopic effects from historical images we were never “meant” to see in 3D. In this post, I’ve described and demonstrated a handful of techniques that have occurred to me so far along these lines. Can you think of others?