In this second edition of what is rapidly becoming the classic text in the field, the authors repeatedly show their dedication to research in related topics, and accept the responsibility for maintaining synchronicity with the expansion of the field. This book is meant both as a survey of approaches for the practitioner, as well as a textbook for the academic. The material presented is self-contained, ranging from underlying mathematical and physical properties to subsequent algorithmic and formula development.
The book begins with a discussion of realistic expectations from image synthesis. Image synthesis provides a framework for constructing new images based on the understanding of the physics, biology, and chemistry of natural phenomena and realistic scenes. Typical applications include standard computer graphics imagery based on two- or three-dimensional model descriptions. Other applications involve images containing randomly assigned pixel values where the overall structures depicted follow realistic constraints. Image synthesis is also used as a verification step for testing image processing operations by analyzing the resultant image for consistency.
The overall structure of the book is generally the same as the first edition (in terms of chapter and section headings), but there are a number of important differences. First, the current edition has 33 more pages than the first edition. Extra exercises seem to be added to many sections. While section 8.2 (“Image Display and Human Perception”) was downsized, more material was added to section 8.3 (“Fast Global Illumination”).
Two sections were added: sections 5.5 (“Environment Map Illumination”) and 7.8 (“Lightcuts and Multidimensional Lightcuts”). Both present exciting areas of illumination research. Environment maps provide distribution parameters of illumination within the covered environment. By casting and probing rays into the environment map, illumination sampling can be computed in an efficient manner, even when the rays do not intersect a particular object geometry. Lightcuts provides for a comprehensive and scalable approach to scene illumination. This model integrates arbitrary geometries, many light sources (pointwise and area), direct and indirect illumination, and diffuse and nondiffuse materials into a tree structure that can approximate a large number of point lights utilizing a relatively small number of rays.
Chapter 2 concentrates on the physics of light transport. This includes radiometry and light emission calculations and the interaction of light with surfaces. Rendering and light measurement equations are developed. Chapter 3 reviews probability (sampling random variables and variance reduction) and Monte Carlo theory. The authors explain why Monte Carlo techniques are useful, and apply the theory to calculate integrals. Chapter 4 advances the discussion of computing light transport by including adjoint equations. Using the potential equation with the rendering equation forms an adjoint system of equations, which provides a computational methodology for illumination based on geometric optics. The global reflectance distribution function is also calculated, along with path formulations.
Chapters 5 and 6 apply the probabilistic theory developed in chapter 3 for stochastic path-tracing algorithms. Chapter 5 begins with simple ray tracing (deterministic and stochastic) and illumination (direct and based on environment maps). Then, indirect illumination approaches are provided for general light tracing algorithms. Chapter 6 investigates different aspects of radiosity. Radiosity emanated from heat transfer research 50 years ago, and currently is applied to rendering and global illumination for scenes containing surfaces with generally diffuse materials.
For radiosity, the finite element method is typically used to solve the rendering equation. These methods differ from the above-mentioned Monte Carlo algorithms that incorporate all types of light paths; instead, radiosity methods process paths that leave a light source that is reflected by the diffuse surfaces in the scene. This chapter then considers relaxation and discrete random walk methods for radiosity. Photon density estimation methods, variance reduction of sampling methods, and hierarchical refinement for clusters are also discussed. The final chapter, chapter 7, combines prior methods, and formulates them as hybrid algorithms. Among these are multipass methods, bidirectional tracing, and photon mapping.
The text is a solid and comprehensive treatment of advanced global illumination methods. The presentations are straightforward, mathematically and physically; relevant exercises are provided; and ample illustrations are used to elucidate the text. The reader should have a fair dose of mathematics and physics background to make the most of the material presented, but an undergraduate education in these subjects should suffice. The authors have continued their tradition of bringing forth a serious treatise on the subject of advanced global illumination.
