27 July 2012

3 Point Poser

You're familiar with how, in a strong upshot or downshot, the verticals of buildings "fan out" or converge, rather than being parallel. We see this in 3-point and 2-point-vertical views, as well as fisheye. Those verticals converge upon the VP that we're most nearly looking at, and widen on the opposite end of the picture. You can see this effect dynamically by panning around within Street View in Google Maps or Google Earth: One end of the picture squishes and one broadens as we mouse to and fro, the squishing happening where a VP comes near our frame of view.

So why is the opposite happening in this picture, made long before Photoshop made such distortions easy? The horizon isn't high in the frame (the view seems almost as much upshot as downshot) and the buildings and cable car which we are necessarily looking up at--given their appearing above the horizon--broaden at the top like thunderclouds. Weird, right?
Archives San Francisco Chronicle

Shouldn't those verticals be, if anything, leaning in toward the top?

I have two hypotheses, one simple, one conjectural:


I think what happened here was that the photog shot with a short (wide-angle) lens aimed down, putting his center of vision on the ground, not among the Santas' bellies but their boots. Then the bottom of the picture was cropped off. Just that simple. The horizon was in fact high in the frame, but the crop conceals that.

BTW, we don't see things like this this thunderhead effect happening with our eyes, because our peripheral vision is weak and lacks detail.

But in cameras the image receiving surface (where film once lay) is flat. So whatever perspective logic is dictated by the lens setup -- in relation to the physical scene -- prevails over the whole flat receiving surface. Point a wide-angle camera at the ground a few feet away, and the verticals converge toward the bottom and widen at the top, consistently, mathematically. There is in this case no muting of the effect: no fisheye to curve the straight lines, or imperfect spherical retina to lose details. The resulting perspective prevails both above and below the horizon, as all perspective does. That basic fact is one of the first things beginning students either grasp, or get caught up short on.



The second hypothesis is worthy mostly just because it leads to a movie that feels like a tiny tropical vacation (You'll see.), but I advance it seriously:

It is that some cameras allow the user to tip the receiving surface. That could create this thunderhead effect. If the surface was tipped to make the image of the building tops have to travel a tiny bit farther in the camera's inner space, they would spread bigger.


The same could be accomplished with the now-quaint optical print-making device known as the enlarger. The challenge in either method would be to not have major parts of the image go out of focus. This could probably be managed by using smaller aperture and longer exposure.

The thunderhead effect may have been in vogue in news photography, c. late 60s -70s - ? Yes, to judge by the collection of vintage SF Chronicle cable-cars photos I got this from, which contained at least one other example of similarly counter-logical perspective. Shooting this way adds jazzy, energizing diagonals, something I'm forever telling comics students and myself to do, to avoid dull drawings.

So how do I know that tipping that receiving surface risks out-of-focus results? Because that very adaptation, called tilt shift, can be used to create an artificially abbreviated depth of field. This is one of the several tricks used to accomplish this bizarre and utterly singular effect (see below), which you may also have seen in insurance commercials on TV. The other tricks are heightened color saturation and sped-up motion. Uncanny how they combine to create a world that looks real and toy at the same time, isn't it?