--------------------------------------------------------------
                Quaternion toolbox for Matlab
                http://qtfm.sourceforge.net/
       Mailing list: qtfm-announce@lists.sourceforge.net
--------------------------------------------------------------
Copyright (©) Stephen J. Sangwine        (S.Sangwine@IEEE.org)
2005, 2006    Nicolas Le Bihan  (nicolas.le-bihan@lis.inpg.fr)
--------------------------------------------------------------

                     Release Notes

Version 1.0 14 December 2006

This release does not mark a major update to the toolbox, but
reflects the fact that the previous version has now been in use
for long enough for problems to come to light.  The toolbox is
now sufficiently mature to have a version number that does not
start with zero.

We have included with this release a new function svdj() which
computes a singular value decomposition using a cyclic Jacobi
algorithm. It will work on real and complex matrices as well
as quaternion matrices. This code therefore demonstrates some
of the tricks needed to write code that will work on real and
complex matrices as well as quaternion matrices. We believe this
may be a useful test feature.

The Matlab functions ceil, floor, fix and round have been
overloaded for quaternion matrices.

Version 0.9 13 July 2006

The Matlab constructor function complex() has been overloaded
for quaternions, making it possible to write complex(a, b),
where one or both of a and b is a quaternion.

The two Matlab functions conv and conv2 (convolution) have
been overloaded for quaternions.

A new function convert has been introduced to replace the
previous functions double, single, uint8 etc. convert works
on the elements of a quaternion, e.g. p = convert(q, 'single')
results in a quaternion with single precision components.
The previous functions have been removed as they cause Matlab
to execute a statement like X(1,2) = quaternion(1,2,3,4) even
though X is of type double - the result is a corrupted element
in X, and no error is raised.

The private function isoquaternion has been modified to work
correctly when the supplied complex number has zero imaginary
part. This fixes a problem where log(quaternion(1,0,0,0))
failed instead of returning 0.

A number of minor bug fixes have been incorporated.

Version 0.8 24 March 2006

With this release we have made available the det, svd and eig
functions. There are some limitations in these functions (see
the source code) which may be removed in later releases.

All of the Fourier transform functions now work for complex
quaternions as well as real quaternions. A transform with
real data and a real axis will be computed using a quaternion
transform. If the data and/or axis is complex, a complex
quaternion transform will be computed.

The following functions which were previously unimplemented
are now implemented and work for real and complex quaternions:
log, sin, cos, tan, sinh, cosh, tanh.

The function power.m has been implemented, making it possible
to raise a quaternion to a power using the .^ notation. This
function uses logarithms in the general case, but implements
certain special cases such as powers of -1, 0, 1, 2, and 0.5
using more direct calculation. The square root function has
also been implemented.

The functions isreal and sign have been added.

The following functions now work correctly for complex as well
as real quaternions: unit, angle, axis, exp, change_basis,
orthormal_basis.

A directory of test code has been added. There is a script
called test.m which runs all the test code, and various
scripts for testing parts of the library. The test code is
not comprehensive and will be expanded in due course.

Structure indexing has been added (in a very limited way).
See the function subsref.m for details (work in progress).

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

Version 0.7 27 July 2005

This version is the first released on Sourceforge.