This README written by John Barron, modified by Michael Hirsch

This program, lucas.c, is an implementation of Lucas and Kanade 1981, 1984
with some modifications proposed by Simoncelli, Adelson and Heeger 1991.
To compile:

gcc -O3 LUCAS/lucas.c -ISOFTWARE -o lucas -lm

Everything is in a single file so a make is really not necessary but...
This program pre-smooths the image sequence with a spatio-temporal
Gaussian and uses 4-point central difference. Hence, if the standard
deviation of the Gaussian is 1.5 you need 15 images. Examples of run
files are also present. (You'll have to change the directory names where
the input and output are directed.) You just specify the central image
(I used 9 always) and the program figures out how many images to the
left and right it needs. The program also use confidence measures based
on Simoncelli, Adelson and Heeger, I used 1.0 typically (0.0 means
no thresholding). To see the programs full range of options just type
lucas on your terminal. An example run:

lucas newbinarytreet. 1.5 20 1.0 /data/images1/BURKITT/testdata /data/images1/BURKITT -B 150 150 -C /data/images1/BURKITT/outdata/correct_trans20

means run the program with the translating tree sequence, smooth with a Gaussian
of 1.5, compute flow for image 20, use a threshold of 1.0, get the input image
data from directory /data/images1/BURKITT/testdata, put the smoothed data in
the same directory and put the output in /data/images1/BURKITT. Natually
you'll have to change these names to the appropriate directories on your machine.
-C indicates the correct answer is available and that you want error analysis
conducted and that data is in /data/images1/BURKITT/outdata/correct_trans20 (change 
this as well as appropriate). Note that in addition to thresholding on the
eigenvalues the program also does not compute velocity where the determinant
of the appropriate system of equations is singular. The program is rather verbose so 
you might want to direct the output to another file. The -T option allows you
to threshold the raw normal velocities (in the lucas.*R* files) with a spatial
gradient magnitude. [So there are two types of normal velocities, which we
call Raw and Least squares normal velocities.] The -S option allows you to
print out the smoothed images. The -L option (default) is for standard
Lucas and Kanade with eigenvalue thresholding and 4 point central differences.
The -M option computes the flow using Sinoncelli (CVPR 91) et al.'s
modifications.
The program is reasonable fast (on a Sparc 2); I'd say about 1 minutes for most
sequances.
See the README above for information about the output format for flows and use 
software in ../SOFTWARE to print out the flow fields in postscript or to
do error analysis.
-------------------------------------------------------------------
[The program has been changed (as of early March 93). Originally the
program was written to accept smoothed image data. The smoothing proggram
was then incorporated into horn.c as a function, the smoothed data
was computed as unsigned char. The new version computed the smoothed
data as float and usually produces better results.]
-------------------------------------------------------------------