« Killing a Work of Art | Main | VR Objects »
December 08, 2009
Panoramas Vs GigaPans : Jared as Creative Director of New Media
I've been meaning to write on this topic for sometime and today I read something written by my friend Alan and now I know it is time.
Alan - you inspire me.
So he recently posted to his blog an entry about
GigaPan Education and Research Incentive.
The GigaPan is an ingenious device for shooting panoramic images at ultra high resolution. It is a robotic device designed to work with almost any camera and to be universal it uses a finger to press the shutter button. Its a great device.
And yes I own one.
But I have a problem with GigaPans and I want to share it.
Before I start to ramble I want to say that I don't want to take away from Alan or the work people are doing with these devices - I just have one problem.
To begin let me go back - to my first digital camera - a Casio QV10. I loved that camera. 320x240 resolution, internal memory, no flash. I paid $200 for it and a year later sold it someone else for $200. I replaced it with the QV100 which had a "Panorama Mode". When you were in "Panorama Mode" it would leave move a strip of the last picture you took to one side of the view screen so you could align it and make panoramic images. They also gave you software for stitching/morphing. This was my entry into panoramic imaging.
Its been a wild ride since then.
I've also worked with panoramic film cameras like the Horizon 202 or the Widelux both of which take a longer than normal 35mm negative. A real bear to print (and even to scan).
I learned a lot about panoramas.
I learned that if you shoot with the camera vertical you get "more" panorama (and need more photos).
I learned that panoramas of vistas are easier than panoramas of rooms (though I wouldn't learn why for years).
I learned that you need a lot of overlap if you want morphing software to do a good job.
Then I learned about Quicktime VR.
QuickTime VR was more than just a panoramic viewer. It changed the game - it addressed the fact that the panorama was more than a photo - it was data that could be run through a viewer to give an experience.
The panos that I have been taking had pretty much been a single strip and if mapped to the inside of a cylinder that the experience could start to mimic looking around the actual space.
The next step with QuickTime VR was cubic VR wherein the entire 360 degrees (in all directions) was mapped onto a cube (giving the sensation of a sphere) and allowing people to spin around a point in space.
I became obsessed with the various methods in creating spherical or cubic VR. The best way was with a fisheye lens using 2 (though generally 3) images that were unwarped and stitched together. This pursuit became mired down by the litigation by IPIX who claimed that they patented the mathematics surrounding the unwarping of hemispherical lenses.
A few years ago IPIX went bankrupt but not before stifling a very active community. Their software was built on a token model where you had to pay for every panorama you made!
While I explore the other tools out there and continued to work with fisheye lenses I learned a few new things.
1) You have to rotate your camera - not around your body - but around the optical center - or the nodal point of the lens. Tripod heads were designed for various cameras and one of the best companies out there for them was Kaidan - who recently went bankrupt.
2) I also learned that you need a lot of resolution. So as the cameras got better so did the panoramas.
At this point I started to work with the Nikon Coolpix line of cameras. You could use true "spot" fisheye lenses with them. Over the years I have used these same lenses with 6 different models of Nikon cameras every time increasing the resolution of my camera.
My most recent iteration is the Nikon P5100 point and shoot camera. Don't laugh this little camera boasts 12 Megapixels and can shoot at an ISO of 50.
With Kaidan gone I have purchased from some Italian panoramic companies - but they have all been very camera specific. I recently picked up this very nice wooden bracket that supports the lens only (so the camera is behind it).
But I digress.
Over the years I have worked harder and harder to find high resolution cameras with the widest field of view possible to create a full spherical image. These images are 360 degrees around and 180 from top to bottom - basically you remap a sphere to a 2:1 rectangle (the poles get distorted but the data is there).
This type of image looks like this:
FYI this image was taken without a tripod - instead I used the virtual tripod method which incorporates a bubble level and a plumb bob.
But the goal is not to just create a cool looking panoramic image - it is to experience that image as a "point in space".
If you have Java installed on your system then this is a navigable panorama. Click inside and pan around - look up - look down.
And, yes you can also zoom in and out using the + and - keys on your keyboard. I generally don't encourage that though and this is where me and the Gigapan don't always see eye to eye.
The experience of a panorama comes best not from zooming in or out but from panning around. The most amateur mistake when displaying QuickTime VR is too set your initial zoom as too wide. The trick is to get the field of view to mimic a normal viewing angle and then allowing every point around the sphere to feel natural and give the sensation of looking around a point.
Remember - with your eye you can pan and tilt but you can't zoom. Your eye only sees what it sees roughly the equivalent of a 50mm lens of a 35mm camera (or a 35mm lens on an APS DSLR).
From a data perspective it is really east to see how quickly the 2:1 equarectilinear mapping of a sphere becomes a standard as all the data can be placed on it regardless of acquisition.
Years ago I saw a spherical projector globe and was able to get my own images on it because its just the obvious choice for data.
Creating repositories or collections of panoramas or even panoramic data sets is a great idea educationally and historically and I am all for what they represent in terms of both time and space.
So this brings me to my trouble.
The Gigapan takes the opposite approach to shooting panoramas. They zoom the lens in as close as possible to take as many pictures as possible to create this mega high resolution photograph.
While the results are cool and great to zoom in on I always get upset because I find myself panning around the photograph. Not panning around the space. The viewer itself is more concerned with the zooming in and out capabilities that giving me that feeling of immersion and to me it is a step backward.
Don't get me wrong - I love what it does - but I really wish there was a viewer that let me (and even if it was only at the one 'normal' field of view pan around in space instead of pan around the photo.
If there is a viewer like that please let me know.
Then encourage others to use it as well so we can experience those spaces that Alan and others have spent so much time and energy to capture and share.
Some of my panorama work is located here:
http://fc.case.edu/newmedia/pano.html
It is funny because every few years I find new technology which does a better job of stitching and rendering and I am tempted to go back and rerender it all.
I save everything.
I have to go hunt down that old Casio.
Posted by jeb2 at December 8, 2009 07:20 PM
Trackback Pings
TrackBack URL for this entry:
http://blog.case.edu/jaredjared/mt-tb.cgi/21644