Thomas K Sharpless is well known in the panoramic world as one of the true pioneers of panoramic and VR photography. Thomas has an extensive background in computer programming and he has contributed a number of techniques and ideas to the panoramic photography community, most notably the “Panini” projection.
Thomas has been a stereoscopic panoramas evangelist since 2014 and he has evangelized us too. 360Cities, with Thomas’ support, is pleased to launch stereoscopic panoramas – you can read more about stereoscopic panoramas on 360Cities here (link to help guide)
-Please can you briefly describe what 3D or stereoscopic panoramas are and how they are created?
The right word for this type of photo is “stereoscopic”. “3D” should be reserved for objects or spaces that really have 3 dimensions, not for images (though that misuse is very common).
A stereo photo is a pair of images, taken simultaneously with two lenses placed like our eyes, about 65 mm apart and looking in the same direction. When presented to the two eyes by a stereoscope, these images give most people the impression of seeing a 3D space. A stereo panorama is a pair of 360 degree images, which when viewed with synchronized pano viewers presents a stereo pair. The most popular stereoscope for viewing stereo panoramas is a virtual reality headset.
Stereoscopic depth perception results from the brain comparing small shifts of position, called discrepancies, between the left and right images. It is important to know that this process depends very strongly on the discrepancies of fine details and textures, as well as of the perceived outlines of objects. And that about one person in 6 has little or no stereo depth perception.
The most obvious way to make a stereo panorama is to stitch together a series of stereo photos, taken by turning a pair of cameras. But such photos will not easily stitch to two seamless spheres, because each series was taken from a moving viewpoint, not from a single fixed point as assumed by standard stitching software.
There are two basic ways to handle this problem:
- Take a large number of closely spaced stereo views. Then the panorama contains just a narrow vertical strip from each photo. If they are narrow enough, the errors between adjacent strips will be too small to see. This method is very reliable but has problems with moving subjects.
- Take a small number of stereo views, as for a regular panorama, and hide the errors by careful composition and masking. Moving subjects can be handled by masking, as in normal panography.
Both of these methods have single-camera variants, that need a really static subject because the left and right views are taken at different times.
In any case, the stitching process must faithfully preserve the stereo discrepancies captured by the cameras. That means the left and right images of each stereo pair must be aligned and warped the same; any variation will lead to “muddy” stereo views. This matching requires special stitching techniques. My PT3D software makes it easier to achieve.
-When and why did you begin to create stereoscopic panoramas?
I have long wanted to make stereo panoramas in forests, because it is so hard to see the space between the trees in 2D photos. That is still a largely unrealized project.
I made my first stereo panoramas in late 2014, with the single-camera, many-views method. At that time I was involved in a project using photogrammetry to build 3D models from sweep panoramas taken with iPhones, and I also experimented with rendering those as stereo panoramas.
By mid-2015, like several other panographers, I was using a stereo pair of small mirrorless cameras and the few-views, clever-stitching approach, which suited my preference for shooting ‘live’ scenes from a monopod. The great difficulty of that method has led me to develop software that helps PTGui align and warp such panoramas much better than it can do on its own.
-Why do you think stereoscopic panoramas are so compelling? (this might be addressed in the question above)
Much more than conventional photos, 360 panoramas are about space. So it is really satisfying to be able to ‘see’ the space in a stereo panorama. This greatly enhances the sense of presence in a VR viewing situation, even without the full motion-parallax provided by a 3D model. I have seen two people, when putting on a headset showing a stereo view of a big church, start walking forward — a sure sign of “presence”. That never happened with a 2D version of the same image.
-How many photographers do you believe are currently creating stereoscopic panoramas and how large do you expect the community can become as VR adoption increases?
I know of about a dozen commercial panographers who produce stereo panoramas, and would guess that there are at least as many more unknown to me. The facebook group “3D Stereo Panoramas” currently has 1,408 members. More than 30 people have asked to be beta testers of PT3D, which will go commercial next month. Its sales figures may tell the story better. But in any case it is clear that interest in stereo panography is growing quite fast.
A big thanks to Thomas for this interview and his support.