NOTATION

The Toolbox functions follow a number of conventions.

For time series data such as a trajectory, rows correspond to increasing time.

For position data, columns correspond to points.  The matrix has 2 rows for 2D data
such as image coordinates, and 3 rows for homogeneous image plane data or 3D points,
and 4 rows for homogeneous 3D data.

For functions with a complex number of options the common function tb_optparse is
used.  This allows a variable number of name/value argument pairs, and also supports
simple enumeration, and loading of arguments from a structure rather than name/value
pairs.  All functions that use tb_optparse support a verbose switch that prints more
detail as the function executes.

This version of the Toolbox makes use of many object classes, and particularly
vectors of objects to represent things such as image feature points or lines.
Often we wish to obtain a vector of scalar properties from the object vector, but 
Matlab returns a comma separated list in this case.  We therefore adopt the convention
of naming the object properties with a trailing underscore, and creating methods
that return the property in vector form. For example, the method u() returns the
value of the property u_.

Most objects in the Toolboxes are derived from the handle class, which make it possible
for methods to change the state of the object in a convenient way, that is, writing

x.modify()

rather than

x = x.modify()

This introduces danger because

y = x;
y.modify()

will change both x and y.

Most object constructors accept an object or object vector and return a copy

y = Object(x);

All objects have a char() and display() method which provides a succinct summary of 
the object state.

Imported packages live in the directory /private below the appropriate Toolbox or
within a class directory.