A Seriously Time-Consuming Project!

In which We plumb the depths of our creative stubborness

Last year I started trying to shoot in 3D with the intention of printing them out on, or at least in front of, lenticular screens. Lenticular screens are typically plastic sheets with vertically oriented half-lenses, that is they warp light from side to side but not up and down, which can be used to view prints that have multiple images that either present motion or a combination of false depth and true stereo, or both depth and motion.

The idea is that each eye sees a slightly different strip of the area beneath each lenticule, of which there can be between 16 and 80 per inch laid parallel across the image. And the magnification factor of the lens is such that many individual strips may lie between each lenticule which can create an additional 3D effect by showing different views as you turn the print or move to the side, which in turn can simulate the views you actually see in real life, or in a true hologram. You've probably encountered this material on DVD packaging where the image changes depending on the angle at which you hold it.

I started off with a cheap prism beam splitter that combined two images from approximately the distance between human eyes into a single frame. It was clearly 3D stereo but had rather poor optical qualities as far as focus, vignette, exposure and consistency.

So I eventually bought a second D2X and matching zoom and prime lenses and connected them with a synch cable and was able to take some fairly good stereo pairs.

Now lenticulars are normally made with many individual images spaced close together, either by using a single camera on a graduated track to take many exposures, or a series of prisms and/or mirrors to take many views in a single exposure, or just lots of small (because they must be close together) cameras. But, of course, I want to take the pictures underwater, where you can't use multiple exposures because nothing can hold still, can't make a complicated optical system practical, and can't synch the many small cameras needed for motion capture.

Two cameras was about all I thought could be reasonably made to work. So I started working on a software solution ...

Eight months later I'm ready to start over again: I've had incredible problems getting the two images properly synchronized. Not only the exposure but the focus point, the color response and the depth of field all need to be exactly the same. Modern cameras just don't work that way: each frame is uniquely focused and exposed depending on the conditions impinging on the cameras precise location. The world is different (enough) one foot to the side.

So I've finally traded in my slightly mismatched pair of D2Xes for a pair of identical twin D3s with sequential serial numbers. And I'm finally getting consistent color response and exposure from identical manual settings but finding out that the switch to the larger sensor has reduced my depth of field to 4 inches! Fortunately, the large sensor is much less noisy and I can stop down several stops and still get a better image than from the D2X with aperture wide open.

I'm very interested in hearing from anyone who knows of software that can take hyperstereo pairs and produce intermediate views; morphing based on outlines does not qualify. The lenticular print shops that I've found do not have anything that can do it; one of them is rather desperate to see what my software can do, presuming I can make it work. But if it's already out there, I might as well just buy it ...

More reports as events warrant.

-alberich

P. S. How to view crossview images

If 3D images are presented normally with the left view on the left side and the right view as expected on the right side, someone with good control over their eyes (which generally means someone with bad eyesight like myself) can stare "through" the images and make them line up such that the 3D effect can be perceived. But this will only work when the corresponding features of the two images are closer than the distance between the viewer's eyes, ~3 inches, which means the images have to be correspondingly small. This is a consequence of the simple fact that the world never requires us to angle our eyes away from each other to see it.

If the two images are reversed horizontally then you must look slightly crosseyed at them in order for the right eye to look at the left side while the left eye sees the right. While this may be less obvious than the previous method it has the the advantage that the images may be much larger, the only limit being how much you can stand to have your eyes crossed. I typically setup two screens and can comfortably view images at 1280x1024 for editing.

One trick that may help will be to hold your finger in front of your face and look at it while paying attention to the images behind it. When you can see three images, one 3D in the middle and two extraneous on either side, you can move your finger down and focus on the crossed pair. Usually your eyes will snap to the 3D image and maintain it as this is what your visual system is designed to do.

Getting closer to the images makes it substantially harder as you must cross your eyes more and you may actually be able to feel the strain in your eyes. The easiest way to view crossed stereo pairs is to bring up the largest image size available and then move back from them so that your eyes need only slightly cross to view them.