A University in 3D

The University of Mississippi embarked on an ambitious project to create both 3D video and stills. Robert Jordan had to learn quickly how to create and reproduce 3D imagery. This is his story:

At the time there were no affordable 3D digital still or video cameras commercially available, so Matt and I each cobbled together a rig so we could capture Ole Miss sports in 3D. Ideally, the lenses on a 3D camera should be about 2.75” or eye-width apart. Matthew’s video was to be the centerpiece with my photography playing a supporting role in the Ole Miss 3D experience. We each started on different, but similar paths, but ironically ended up coming to the same conclusions and solutions in the end.

Matthew ordered two Canon 5Ds cameras and a 3DFilmFactory 3D-BS Mini beam-splitter to shoot HD video with. The 3D-BS consist of a box with a 50/50- mirror mounted at 45-degrees. The mirror allows 50% of the light hitting it to be reflected and 50% of the light that hits it to pass through the mirror. One camera was mounted behind the mirror box shooting through the 50/50 mirror while the other was mounted to the top of the mirror box above looking down on the 50/50 mirror. The cameras both ‘see’ whatever is in front of the splitter mirror, with one camera shifted off center horizontally to achieve a ‘left’ and ‘right’ perspective.

I decided to take a simpler approach and set off to build a rig to mount two Nikon
DSLRs side-by-side. I had access to pair of Nikon D40 cameras, so I mounted
 them on a flat aluminum bar and I shot some test shots around campus. Unfortunately, the D40 does not have any external shutter release and I was not able fire both
 cameras in perfect sync. The only other cameras I had were very heavy and large Nikon D3 bodies. Fortunately, I discovered that if I simply connected the two D3 10-pin connecters together, pressing either camera’s shutter button fired both cameras in perfect sync. However, the cameras were far too heavy for the simple aluminum bar I was using and the lenses were too far apart with the cameras side-by-side. Inverting one of the cameras would get the lenses much closer together so I contacted Physics Machine Shop Supervisor Mike Reep to see if he could build a better mount for me. Mike took some measurements after I showed him how I wanted the cameras to mount and called me a couple of days later to see what he had built.

Mike is an artist with metal and CNC milled a beautiful, sturdy rig from billet aluminum that exceeded my expectations. The rig is a 15”x8”x3” 4-sided box that holds both Nikon D3 bodies, with one camera inverted so the two camera lenses are just a touch over 5” apart. The generally accepted rule of thumb is to multiply the optical distance between the two lenses by 30 to arrive at the closest distance a subject could be from the camera: 5’ x 30 = 150” or about 12.5’.

I mounted two D3 bodies with identical lenses and reset all settings to the factory defaults and took the rig for a walk around the campus and shot various buildings and people being very careful to set both camera bodies to the same ISO, aperture, shutter speed, focal length and focusing distance.

Once back in the office, I downloaded the images and processed the files in Photoshop. The first step is to rotate the inverted camera images 180-degrees and then render the images for anaglyph 3D. There are tons of great tutorials on this, but basically you copy and paste one image over it’s mate, align the images at the point of focus, drop all the red from the ‘right’ channel and all the blue and green from the ‘left’ channel. Then change the top image layer from ‘normal’ to ‘screen’, flatten the image and enjoy.

Matthew was shooting video with his Canon 5D cameras and performing the same steps in post to render his video in anaglyph 3D. We both hit some home runs that made it to the final product, but also struck out on many attempts.

I discovered a major setback when I sent my anaglyph images to a color printer and then viewed the printouts with the anaglyph glasses. Converting the RGB to the printer’s CYMK ruined the 3D effect on many of the images. I’ve seen anaglyph images printed in magazines, so I knew it was possible. I asked one of our graphic designers to find out what we had to do to make the images work with CMYK inks.

I had other problems to work on, like our school colors; red and blue, which are very close to the colors of the anaglyph glass lenses. Whenever either red or blue appear in an image, your eyes fight for dominance and the 3D effect is spoiled. Fortunately, I found that desaturating the offending color before post processing in Photoshop solves the problem.
Matthew and I were talking one day and he said he had discovered a tutorial on the Internet for making anaglyph 3D images from a single photo. The process involves dividing the image into different planes based on the distance from the camera and creating a depth mask for the different planes. Once the depth mask was created the red/cyan colors in the image were shifted in different directions on either side of the focus point of the image and varying amounts depending on the distance from the focal point.

I used the process on some images and was blown away at how well it worked. If the depth mask is done crudely, the 3D effect was not very good, but more intricate and shaded depth mask resulted in some very impressive anaglyph images. Plus the faux 3D images looked much sharper when viewed without the glasses and they printed to CMYK without any loss in the 3D effect!

Robert Jordan is the Director of Brand Photography Services at The University of Mississippi, and can be reached at rjordan@olemiss.edu