Author Topic: (light field) camera not light (field camera)  (Read 1309 times)

Jack Johnson

  • Sheet Film
  • ****
  • Posts: 667
    • Me on Flickr
(light field) camera not light (field camera)
« on: November 28, 2011, 12:03:55 AM »
Every time I hear someone say they really want a light field camera, I immediately think, oh, like a Tachihara? Me too!

But no, they're talking about integral imaging. If both light field and integral imaging are terms new to you, join the club. The gist of it is some folks are marketing a digital camera that can capture the vector information as well as the color and intensity of the light, and can then reconstruct the image spatially. Normally this is pitched as allowing you to focus after the fact. Yep, read that again. You could also alter the bokeh, or (in this modern age) present a three-dimensional representation of the capture to the viewer. Other than archaeological or forensic work, I have trouble finding the practical value, but definitely something for the cabinet of curiosities, right?

It turns out the ideas behind all of this are almost as old as photography itself. I stumbled across this article that resurects hundred-year-old ideas about leveraging an array of lenses (in this case, pinholes) to capture an array of distinct views (think LOMO Action Sampler on steroids), then using the resulting array to reconstruct the light field.

Of course, a hundred years ago they weren't simulating a light field. It turns out, you can use what's called a fly's eye array to reintegrate those images into a composite view, and like a holograph your eyes will be able to focus on near and far elements in the, uh, image...capture...whatever you want to call the result. Clearly, not traditional photography.

The authors of the article were extremely clever, though. They used a digital back on a medium-format camera, slapped an array of pinholes over the sensor, and voila. I definitely did not understand the math regarding how the pinhole(s) between the lens and the sensor are able to resolve the image, but I believe them, and it occurred to me that you could probably cheat all the way around, ditch the digital gear, and reconstruct an integral image completely analog.

It seems like unlike big sheet lenticular lenses that you might find in a children's toy, fly's eye lens arrays might be hard to come by. You can get them for about $20 per square inch on your auction site of choice, but unless you want to work with an 8x8 array of 3 mm x 3 mm pinhole images I think that's probably a non-starter. Initially, I thought if you took something like two Action Samplers, shot slide film with one, you might be able to butcher the other into something like a ViewMaster that would let you reconstruct the four views into a low-fi light field.

But, if you've already got a hybrid workflow going, and you're shooting 4x5 or larger, you could probably replicate the trick and put a pinhole array in front of the film holder, scan the entire sheet and spend a few hours chunking the results into...what? The world's best wiggle GIF? Whether analog or digital, presentation of the results is the real stumper for me. It seems like if you could get your hands on a 4x5 or 8x10 fly's eye array, you could measure the lens distance, build the pinhole array to match, then shoot transparencies and light them from behind with the fly's eye array in front. The right subject with the right film, I think the result would likely be much more dramatic -- and interesting -- than its current digital counterpart. Whether it would be worth the couple hundred dollar investment in fly's eye arrays to experiment at 6x9, I'm not sure. My gut says either it will be mind-blowing or it will be lackluster.

Help me, Francois. You're my only hope. ;)

The other bit of reusable fun from the initial article is that they had the pinhole array printed, which I thought was a clever trick.

Francois

  • Self-Coat
  • *****
  • Posts: 15,552
Re: (light field) camera not light (field camera)
« Reply #1 on: November 28, 2011, 03:58:34 PM »
Well... though to answer that one... especially when all I can think about is the Nimslo camera...

Thing is there are tons of things mixed together. Lightfield techniques as described in the siggraph papers needs digital processing to do the focussing after the fact since they don't actually record an image as such...

Then, there's simulating depth using multiple images... which needs a lenticular screen over the image to prevent one eye from seeing what the other is seeing.

If you want to try it out for cheap, get yourself a Pop9 like camera. It takes 9 identical images through 9 lenses...

There's also Edmund Scientific that sells all kinds of strange optical stuff.

As for doing the wiggling gif from hell, I think that's pretty much the only thing possible without resorting to very complex math...

Hope that answers...
 :-\
Francois

Film is the vinyl record of photography.

DS

  • Peel Apart
  • ***
  • Posts: 332
  • I don't look like my avatar
    • Waffle blag
Re: (light field) camera not light (field camera)
« Reply #2 on: November 29, 2011, 03:56:22 PM »
Well, I imagine you could do this which is way cooler than the *.gif from hell: https://www.lytro.com/living-pictures/283 (click on bits of the image to explore focus). I think scanning and compositing all the images would be a lot of hard work though on an analogue version.

I'm sure I've seen a paper on pre-digital a plenotptic / light field camera. I'll see if i can dig it out

Hope you make this work!