The parallax bar

[T3D Dr. George A. Themelis <DrT-3d@xxxxxxxxxxxxxxxxxxxxxxxxxxx>]

The parallax bar is a useful instrument to measure parallax in a pair of
stereo photographs.  The parallax is needed to calculate the difference
in height (depth) of any two points.  The bar must have originated in aerial
photogrammetry and was later adapted for SEM stereo pairs too.

While you could measure the horizontal separation between any two points 
in one picture and then the other using a _ruler_ and then subtract them 
to obtain the parallax, the parallax bar will let you do the same thing
but more accurately (resolution of 1/100th of a mm or better) because it 
is essentially a long micrometer and takes advantage of stereopsis to do
the measurement.  You must be able to see depth to use the bar!

To use the bar, you first position your stereo pair under the mirror
stereoscope and align it properly.  Good stereoscopes come with a
binocular attachment which allows you to focus on the center at higher
magnification.  The bar in both ends has two pieces of glass and in each
glass a cross or X is engraved at the center.  Adjusting the bar changes
the spacing of the glasses and a micrometer at the end will give you a
reading, indicating the displacement of the glasses.  

The procedure to measure parralax is as follows:  Say that you want to 
measure the height difference between a hill and a nearby valley.  
Adjust the bar so the two X's at the end coincide with the top of the hill. 
The X's are fused and they appear to be floating over the 3d image.  Ajust 
the bar so they are touching your object of interest.  Record the numerical 
indication that the bar provides.  Move and adjust the bar so that the X's 
coincide with the bottom of the hill.  Record the new numberical value.  
Subtract the two.  The difference is the parallax and can be expressed in 
mm.  The absolute values that the bar provides do not mean anything. Only 
the differences count.

Viewing the matching crosses or X's in depth, shows them floating over
your 3d image.  By slightly adjusting the bar, the floating spot appears
to be moving up and down.  You can only work the bar if you can see this
floating spot and its relationship with the points you want to measure.

I used the bar at Northwestern and did a reproducibility test by marking
10 spots and repeating the measurements over and over again.  My results
were consistent.  I tried the same experiment with two lab fellows and
they both gave inconsistent readings.  I believe that it takes good 3d
visual skills and practice to work the parallax bar with accuracy.

If you don't have a parallax bar you could measure distances in photographs
using a microscope with a calibrated stage or a traveling microscope.
Modern computer manipulation of images I bet is changing all that.  Instead
of recording the stereo pair at the recommended 5 to 10 degrees angle of
tilt, one could use a much greater angle difference, 15, 30, 45.  This 
will create much larger parallax and then even a simple ruler can be used
to measure distances with acceptable accuracy but without the use of 
stereopsis.

George Themelis


------------------------------