When I was learning to program there was a textbook, which I still
have, which had a chapter on stereo imaging. It seemed to me that one
could take two images and figure out the original depth of a scene.
A NASA textbook confirmed this with a description of the Mars Viking
Cameras. I was fortunate enough to have had a chance to work with some
of the younger members of this team, prior to their retirement.
When visiting Europe I would often hear the phrase. "That palace was
re-built from photographs." I was to learn that this procedure is
called photogrammetry. A company I worked for used such a process as
for making maps, for input into a CAD system.
This is still pretty much a university graduate program. There are
a lot of papers written about this. While headway is being made in
medical applications practical applications are still problematic.
I noticed the camera-based roll reader when visiting Ruedesheim.
Somewhere on the MMD it was noted that it could read the notes off the
music roll a playing instrument. If the resolution is high enough on
the camera, with modern processing power this is possible.
When I worked for Apple computer I tested the laser printers. These
are capable of putting marks down with a resolution of 1/600th of an
inch. The paper was allows to stretch or be miscut by 2 mm, or 12
pixels. I wrote the tests used to verify this.
Prior to visiting Wendel's Mechanical Music Manufactory I had already
given such things a bit of thought. As I program daily in Postscript,
I decided to write a MIDI to PS program for making simple test rolls
for the 8-note paper organ. This is still a work in progress.
There is a Linux program (for which I made a Macintosh version
http://www.delectra.com/tina) called TINA. This program is a
laboratory for working with images. I have been using it in my study
of the Jaquet-Droz dolls. I have nearly 100 photos of the Musicianne.
The problem I have found is that I have no reference point in which to
align the images. I am still working at this, however I got distracted
by the MMD!
On the other hand. If one was to set up a dedicated system,
calibration targets (which I did not have with me as they are on glass)
and a fixed length focus camera would make an optical roll reader
practical an simple. Such a device can even read barrels or musical
box cylinders. For the latter it would be necessary to control the
lighting. One would of course need more than one picture of the
In working with this software, it seems that PostScript, with its high
resolution and vector mapping, would make an excellent archival format
for rolls and cylinders. What a lot of people do not know is that most
modern Jacquard looms use PostScript to weave patterns into cloth.
It seems to me that .ps or .pdf files would be a good place to store
not only tracker and keyframe scales; these would be a practical way to
store a whole roll. A hole could be defined by a macro. The position
would be as accurate as the system demands.
A water jet cutter would then be more practical than a laser cutter.
While laser is used I understand water jet's are preferred. Of course
the paper would get wet, I think this can be accounted for.
The duplicate roll would be as close to the original as possible. With
a program like TINA distortions in the original could be detected and
If anyone wants to play with the TINA program, I would encourage it.
There is a lot to it, though. I would recommend the Linux version, as
mine is designed to run the stereo re-construction tool. I am still
working on the calibration tool. It runs the demos, but I have had
trouble with data from my uncalibrated cameras. A fixed-focus rigidly
mounted camera is necessary. I find the consumer cameras introduce too