Breakthrough: A 3D Head Scan Using One Camera
So over the last couple of months, Cory & I have been working on putting together an augmented reality short and we’ve been chipping away at all sorts of workflow obstacles to overcome. One of the big challenges is figuring out a way to not have to use 45 cameras because A. it’s freaking expensive & B. it’s really freaking expensive. 45 cameras means 45x lenses, 45x batteries, 45x memory cards, etc. So until we swimming pools full of gold coins like Scrooge McDuck, we have to rely on some clever workarounds.
Stereo Pair Camera Setup
Amazingly, Cory discovered that using stereo pairs when using photogrammetry results in superbly clean meshes. When we captured Jade Bryce, we only had one camera and had Jade hold really still. Cory walked around Jade and took 45 shots that were evenly spaced. A couple of weeks ago, Cory had a stroke of genius and realized that if instead you pair two cameras close to each other, the geometry that’s built from them comes out looking a million times cleaner. So now, instead of trying to photograph our subjects with an evenly spaced pattern of cameras, we’re employing a stereo-pair pattern. We use two pairs of cameras close to each other, similar to how you would film 3D video. Each stereo pair can be evenly spaced although we’re experimenting with what the optimal configuration should be.
But you said you used one camera?
Correct. The way it was faked was by having Cory on a chair while Heather took a series of vertical shots. Then Cory would rotate on the chair and then we would repeat. The camera captures were close enough so that it mimicked the stereo pair setup I was talking about earlier.
What’s most surprising about this setup is how cleanly the geometry comes out using only one camera. Because the person can be on a rotating platform (chair), the acquisition time is cut down from 5 minutes (impossibly long to hold still for that long) to < minute. We’ve taken this idea to build out an automated turntable platform for our shoot day when we capture our characters so I’ll have more information on that later but here’s a picture of what’s going on: