Tuesday, August 20, 2013
Various "Flavors" of Panoramic Photography
Most of us are introduced to the concept of panoramic photography by taking a tripod-mounted camera and rotating the camera taking a series of images that overlap. Software, like Microsoft ICE or other "stitching" software, is then used to calculate a pleasing blend of the images, making use of the overlapped portions of each image. Depending upon a number of factors (including whether the nodal point of the lens was used as the rotational center or not, the focal length of the lens, the distance to the subject, and the proficiency of the stitching algorithm) errors may be found. This type of panoramic photography is called segmented photography (for the image stitching) and also rotational photography (for the fact that the lens must be pointed in a different direction for each segment's exposure. If you think about the film plane you will see that it is curved around the rotational point, but that each segment is made up of a flat image plane (the sensor or the film is not curved).
There is also a "flavor" of panoramic camera that make use of "short rotation". This is used in cameras such as the Widelux, the Noblex, and the Horizon. These cameras have a lens that rotates during the exposure around the camera's rear nodal point and use a curved film plane. Rotating lens cameras produce distortion of straight lines and also must have a limited selection of shutter speeds to work with.
Ignoring "full rotational" panoramic cameras that can do 360 degrees (or more) let's go straight to the "fixed lens panoramic". Examples of this are the Fuji G-617 and GX-617 (for 120/220 film), the Fuji TX-1 and TX-2, and the Hasselblad Xpan cameras (for 35mm). They create a wide aspect ratio image with a single fixed lens and a flat image plane.
The Fotodiox RhinoCam™ (and my Pentax HippoCam) fall into a whole new hybrid category. They are fixed lens designs that also use the segmented (stitched image) approach, but instead of rotating the camera to point the lens in a different direction for each segment, (introducing rotational distortions) these hybrid cameras make use of lens designed for a much larger format and slide the digital sensor across the flat image plane, taking a series of overlapping images that can then be stitched into a larger, rotational-distortion-FREE images.
The secondary benefit of this approach (or maybe the primary benefit, depending upon your point of view) is that you are producing what could only be produced with a much larger (and expensive) digital sensor. Using my Pentax K-5 II as capture camera (a 16 megapixel sensor) and taking a series of 6 images across a 6x7 lens' image plane, the result should be an image in the neighborhood of 80-85 megapixels (image overlap means it is going to be less than 16x6 or 96 MP).