A distorted translation of what is seen by the human eye, dream-like, equirectangular visuals can be created with the help of digital technology. Photo curvature, or pixel bending, is a way of abstracting a panorama or any other picture so that the poles are either pulled in or stretched out across the canvas.
Photographer and videographer, Garret Veley, has shared his common workflow, including the equipment and software used to produce the equirectangular and other re-projections shown. But first, here are a few definitions and examples of the typical types of projections that can be developed.
Equirectangular projection – the way of mapping a sphere onto a rectangle. Think of the way a map of the Earth looks as a rectangle. The north and south poles are stretched way out to fill the frame and as you move towards the equator the distortion gets less and less.
Once you have an equirectangular panorama, many other projections can be obtained with the right tools. Flexify by Flaming Pear is what Veley uses to achieve these advanced projections from the base images.
Stereographic projection – the south pole is pulled in toward the center of the frame and the north pole is wrapped around the outside of the frame.
“Wormhole” stereographic projection – the north pole is pulled into the center and the south pole is wrapped around the outside of the frame.
Other exotic re-projections:
Mercator Cross projection of this base image:
Hyperdouble and Hypertriple projections of this base image:
Composing the beginning equirectangular projection is more work than you might assume. Here is the detailed workflow described by Veley on how he generates his initial equirectangular panoramas:
I shoot with a Canon 5D Mk2. Most of the equirectangular images are shot with a Canon 15mm Fisheye. I have a very well used Nodal Ninja 3 panorama head that I love. For anyone unfamiliar with it, it is a panorama head holds the camera in such a way that it can be rotated around the lens’ No Parallax Point, often called the Nodal Point. I use the Promote Control to keep my hands off the camera, set exposure and control the brackets (typically that’s 3 brackets, -2/0/+2).
With the 5D/15mm camera the number of exposures for one equirectangular is 8×3 around horizontally, 1×3 up and 2×3 down for a total of thirty-three images. Add more if there are people or anything moving around that I need to capture.
I convert the RAW files to 16-bit TIFF with Adobe Camera Raw. I want files that are sturdy and bit-laden as possible to keep it clean during processing. All that warping from the stitcher really stresses parts of the image. The panoramas are assembled in PTgui Pro. I get roughly 14,000 by 7,000 pixel, 16-bit, layered PSB files from PTgui. The PSB file includes an exposure-fused layer and the three stitched bracket layers.
Often, I’ll have it also include every single exposure as separate, individually warped layers. That’s thirty-seven or more layers! It takes a while to stitch and save. The PSB files is usually a monster file, somewhere between 3-8 gigs or more.
In Photoshop I can use the three exposure layers to manually improve the exposure balance of the fused image–shooting a full sphere all around the camera practically guarantees there won’t be just one setting that works best for every direction.
The individual image layers let me adjust the way PTgui has blended the stitched images–inevitably something is moving differently from one exposure set to the next (people, branches, clouds, waves, etc. …) so it’s necessary to patch things up a bit. When I get that done I flatten the file, which shrinks it to a much more manageable size, usually around 700meg.
With the image assembled, I can work on color and contrast. That’s where equirectangular has specific challenges, because the image is horizontally continuous all the way around and vertically it is progressively distorted. You always have to account for the fact that the left side connects to the right side. If you put a grad mask on, it has to be perfectly level horizontally, for example. Any time you feather in a correction you have to watch out for the edges of the frame
So with filters like Topaz Clarity or Adjust, looking at all of the surrounding pixels to make adjustments to any one pixel, you have to have a method to control what happens around the edges of the frame. If all you do is apply the filter globally, it will usually create a sharp transition where the sides meet because the filter doesn’t see the connection around to the other side.
And in an equirectangular image, as you move towards the top or bottom of the frame, the filter sees pixels side-by-side that will not be side-by-side when the projection changes. So what might appear as a subtle correction in one projection can turn into a glaring, sharp-edged error in another. It’s odd stuff.
The approach I use is to pair Topaz with the Offset filter in Photoshop. I make two copies of the equirectangular image and use Offset to slide one layer 180 degrees out of horizontal registration. I apply the same Topaz filter to both layers, then use Offset again to slide the offset layer back into registration. I then grad mask off a vertical strip to erase the horizontal artifacts, and about 10% of the top and bottom of both filtered layers to clean up the top and bottom artifacts.
If it’s a huge correction in Adjust or Clarity, or a really radical filter like Simplify (which I adore!), and it won’t just blend back into the original, then I might have to go for a round-trip back through PTgui to roll the sphere 90 degrees (so the poles are in the center of the rectangle).
When I bring it back into Photoshop, the filter sees less distorted data at the poles and make a clean correction in those areas. I’ll then sent it back to PTgui, unroll it, bring it back into PS and blend in just the poles.
The exotic re-projections of equirectangular images are created using a crazy-cool plugin called Flexify. You feed in one projection, select an output projects, then manipulate a few parameters to find the most interesting degree of distortion within that projection. It’s utterly fascinating.
The best thing about shooting spherical panoramas, especially for landscapes, is that it forces you to engage with your location in a really cool way. Usually when you find a spot that is interesting, you walk around and ask yourself, “What’s the most interesting angle to look at this from?”
But, if you are shooting absolutely everything around the camera the problem becomes, “Where is the best position to optimize my view of this entire place?”
You have to walk around a spot and try to take in all of its features. One must really engage with the entire location. It’s a wonderful thing to do while traveling because you view the same scenery everyone else is viewing…and what everyone else is ignoring–it’s surprising how often you find yourself standing in a spot that you wouldn’t choose otherwise.
About the photographer:
Born in Tulsa, Oklahoma, Garret Veley attended Gonzaga University and currently resides in Bainbridge Island, just across from Seattle. He currently owns a small video production company doing corporate video.
While in college, in 1978 he became a darkroom rat where he learned about the processing side of photography. During that summer he worked as a “hand” on an oil service boat doing acid-frack jobs in the Persian Gulf where he shot around 200 feet of bulk-loaded Plus-X film of service boats, offshore rigs and various scenes around the then-tiny town of Dubai.
Upon the return back to Gonzaga he spent hours of late night processing his backlog of black and white film from Iran and Dubai. Soon he met other photo geeks, most of them being in the Broadcast Television program, which led him down that track too.
“Video production is what pulled me away from still photography–for me, it’s really two different mindsets to shoot stills or motion. These days I make my living in relatively uncomplicated corporate video and to a lesser extend stills. I shoot spherical panoramas mostly for fun.”