[Technical Info]
All these pictures were taken with digital scanning panoramic
cameras that I made myself out of flatbed document scanners and old
35mm SLR lenses. I also wrote the software that operates the
cameras and some specialized image processing software to
handle the rather unusual images they produce.
A scanning digital camera takes images one narrow
vertical line at a time. Each line is really a separate
photograph and there are thousands of lines in a picture. That is
a slow process, so scanning
cameras are only good for stationary subjects under steady
lighting.
My panoramic cameras revolve around a vertical axis to scan an
image of part of the world. The scan can cover any angle up to
365 degrees. The vertical extent of the image (which I also
measure in degrees) is determined by the focal length of the lens and
the length of the CCD image sensor. These cameras can achieve
very
high resolution because the image sensors have large numbers of
pixels, and because the whole picture is taken on the lens
centerline. They produce seamless images
with true panoramic perspectives: cylindrical if a "normal" lens is
used, spherical (technically, equirectangular) with a "fisheye" lens.
My oldest camera is optimized for taking long panoramas at very high
resolution. It has a 10,600 pixel CCD 42mm long, which covers the
full image circle of a standard
35mm lens, at a resolution that only high quality lenses can deliver --
equivalent to 48 megapixels per 35mm frame. I use
Asahi Takumar lenses of 55 and 28 mm focal length and a 16 mm
Zenitar fisheye, which give image heights of 54, 75 and 150 degrees
respectively. I usually scan at 1/2 or 1/3 of the maximum
resolution and record 50 to100 million pixel images, enough for about
a square yard of sharp print. Typical scan
times are 5 to 15 minutes.
A second camera
is optimized for speed with very wide angle lenses, from 8mm to 24mm focal length.
It has a 5,400 pixel CCD 28mm long -- a little
more than the height of a 35mm film frame. With a fisheye
lens tilted up at a steep angle, it can capture images of tall
buildings or the whole dome of the sky (such images need special
processing to correct color registration errors due to the extreme
scanning geometry). A full circle scan with an 8mm
lens has about 36 million pixels and typically
takes 2 to 7
minutes depending on light level. The 16mm lens delivers about
100 raw megapixels. I usually convert these scans to circular
views with about 2/3 as many pixels.
My latest camera is an all purpose compromise with a 5,300 pixel
sensor 37mm long. Its larger pixels give it somewhat better
exposure latitude than its predecessors. There is an article on
that one here,
and some engineering details on the older ones here.
I finish my pictures on a 2 GB PC using
Photoshop , the Helmut Dersch panorama tools, and some programs I wrote myself.
Because of the high pixel density, it is possible to alter viewing
projections fairly radically and still have smooth, crisp images.
I often use projection functions custom made for the image
at hand, especially for architectural subjects. This is analogous
to using the tilts and swings on a view camera, but more
versatile, as one can get projections that no lens could produce.
I make prints on archival and fine art papers with 1440 DPI
Epson printers using Ultrachrome inks. To control
color and tone rendition, I use input ICC profiles specific to
the camera and lighting conditions, and output profiles specific
to the ink and paper. Pixel density ranges from 40,000 per
square inch in the largest prints, to around 300,000 per square
inch (which is more than these printers can really render) in the
smallest. By comparison, you are probably viewing the images on
this site at about 10,000 pixels per square inch.
Many of these pictures contain odd looking blips, often in
primary
colors. That is how the scanning camera renders things that are
moving horizontally -- mainly people and vehicles. Things
get more compressed the faster they move. The coloration is due
to the fact that the red, green and blue images of a given point are
scanned a
fraction of a second apart. I like the way these cameras
change traffic from a major visual blemish to a sort of comical
decoration (obvious in "Twilight Skyline", but try to find the two images of a moving car in "Morris Meadow").