Browse Books

Reviewer: Frederik W. Jansen

Rendering of natural scenes requires the modeling of natural phenomena such as clouds, trees, and water. Complex objects also display many details in their surface texture and color. The large number of these details makes the use of standard graphics input techniques to specify these objects unattractive. A procedural approach offers a good alternative. For most objects and textures, a function or procedure can be written that automatically generates many volume and surface details, and can add the right amount of harnessed chaos that will give the objects their natural and realistic appearance. Conditional and step functions, filter functions, periodic functions (such as sinus and cosinus), and so-called noise functions (functions that add the required randomness) can serve as building blocks for these procedural descriptions. The desired noise function is a pseudorandom low-pass filtered version of white noise. Within the procedural approach, an implicit formulation of this function is needed. An often-used technique is lattice noise, an interpolation of random gradient values located at the nodes of an integer lattice. A drawback of the procedural approach is the difficulty of specifying a function that will generate a given texture or object. Which set of functions, for instance, will generate a suitable family of woodgrain textures__?__ It is hard to analyze a given pattern and come up with a generic function for it. Moreover, the computing-intensive nature of some of the functions makes interactive exploration of the parameter spaces a tiresome exercise. The generation of procedural models and textures therefore largely remains a black art. In this book, some of the pioneers of procedural modeling explain their methods. Darwyn Peachey gives a good introduction to the underlying techniques, the basic functional building blocks, the implementation of the noise function, and filtering techniques for preventing aliasing. Worley, Ebert, and Perlin contribute a wealth of examples taken from their own work, illustrated with pictures and large patches of code. These chapters are straightforward how-to recipes for creating different effects. Musgrave completes the book with three chapters on fractal techniques. The book is a revised version of the authors' original SIGGRAPH tutorial notes. The intended reader is the programmer who wants to know the ins and outs of procedural modeling. In that respect, the book gives a wealth of material and instructions. Unfortunately, the authors did not turn their notes into a real book. The introductory chapter is just another preface. The style, depth, and coverage of the different chapters vary considerably. The book does not provide a systematic and complete coverage of all scientific issues and practical aspects<__?__Pub Caret>. It basically only covers the work of the authors with a lot of overlap between the different contributions. Peachey is the only one who shows that he can write and is able to present the material in a logical order. My advice to the authors would be to ask Peachey to expand his chapter into a book and add the material of the other authors as hands-on exercises. Overall, this is a great book for texture-hackers, but not a balanced and complete introduction to this interesting subject.