Introduced 160 years ago as an attempt to generalize complex numbers to higher dimensions, quaternions are now recognized as one of the most important concepts in modern computer graphics. They offer a powerful way to represent rotations and compared to rotation matrices they use less memory, compose faster, and are naturally suited for efficient interpolation of rotations. Despite this, many practitioners have avoided quaternions because of the mathematics used to understand them, hoping that some day a more intuitive description will be available.
The wait is over. Andrew Hanson’s new book is a fresh perspective on quaternions. The first part of the book focuses on visualizing quaternions to provide the intuition necessary to use them, and includes many illustrative examples to motivate why they are importanta beautiful introduction to those wanting to explore quaternions unencumbered by their mathematical aspects. The second part covers the all-important advanced applications, including quaternion curves, surfaces, and volumes. Finally, for those wanting the full story of the mathematics behind quaternions, there is a gentle introduction to their four-dimensional nature and to Clifford Algebras, the all-encompassing framework for vectors and quaternions.

* Richly illustrated introduction for the developer, scientist, engineer, or student in computer graphics, visualization, or entertainment computing.
* Covers both non-mathematical and mathematical approaches to quaternions.
* Companion website with an assortment of quaternion utilities and sample code, data sets for the book’s illustrations, and Mathematica notebooks with essential algebraic utilities.

Worth a look

This is a very interesting book in the Morgan Kaufmann series, and will appeal to those with a mathematical bent. Visualizing quaternions is broken into three parts. Part 1 treats the elements of quaternions, and parts 2 and 3 treats advanced mathematical topics that place considerably more demands on the reader’s mathematical knowledge (and also on the author).
Part 1 is an introduction for those readers new to the topic. As far as introductions go, it is not too bad. It does in fact contain one important subject - quaternion interpolation - that is not always covered in other texts. Hanson covers interpolation in part 1 and again in part 2. If your interest is computer animation, this may be sufficient reason to acquire the book...analogous to purchasing an album just to get one song. However, if you are completely new to quaternions and want to develop a firm intuition grounded in first principles, then a book that is at least an order of magnitude better is "Quaternions and Rotation Sequences" by J. B. Kuipers.

Parts 2 and 3 are the most interesting parts of the book. Hanson presents a series of small chapters that discuss quaternions from different advanced mathematical viewpoints (differential geometry, group theory, Clifford algebras, octonions). The chapters are small, and so they by necessity contain references to the literature where the considerable background material required for understanding the topics is developed. If you have a good background in differential geometry and some abstract algebra, then the chapters are quite nice. In this sense, parts 2 and 3 of the book are more appropriate for mathematicians.

The technique of including routine, "turn the crank" type of calculations in the text, and deferring the sometimes considerable details and theory to references allows Hanson to cover more topics than usual. However, it is exactly those details that distinguish between what is useful and well conceived mathematical theory from mathematical gibberish. Deferring details to the literature can also encourage an over-reaching of the author beyond his understanding of the material. Hanson has walked a fine line here, but still I must mention two issues that I found annoying:

1) A Riemannian manifold is not specified only by giving the charts ("local patches") as Hanson seems to think on page 352. One must also add constraints on the topology -- typically Hausdorff with a countable basis of open sets. These are not just moot considerations; the topology allows a construction of a partition of unity which in turns guarantees the existence of the Riemannian metric. In particular, the mild condition of paracompactness will ensure the existence of the partition of unity.

2) It is a gross over-simplification, and mathematically non-trivial, to claim the basis vectors of Euclidean space have precise analogs in Fourier transform theory, as Hanson does on page 340. Heuristic analogs...yes... but precise analogs?...only if one has developed the necessary mathematical machinery using the theory of distributions. The inner product relation ei.ej = kronecker delta ij given by Hanson on page 340 would have to be generalized to a delta function. It was one of the major accomplishments of 20th century mathematics that Schwartz was able to put the delta function on a firm mathematical basis with his theory of distributions (for which he received the Fields medal) Before Schwartz, delta functions were at best a useful computational tool in the hands of physicists like Dirac who were guided by their physical intuition, and at worst, an example of the mathematical gibberish alluded to earlier.



In short, this is a good book for those with the mathematical prerequisites. Those with a more traditional background in computer science might be advised to first peruse a copy at their local bookstore to verify it matches their interests.

Beautiful book, decent content

Beautiful production (typesetting, graphics, layout). The mathematics is on the informal, intuitive side. I consider this a luxury purchase, not an essential part of one’s hardcore math library library. Somewhere on the shelf next to Tufte’s books on visualization of data.