The Magic of Computer Graphics: Landmarks in Rendering

The Magic of Computer Graphics: Landmarks in Rendering

Noriko Kurachi, Michael Stark

Language: English

Pages: 428

ISBN: 2:00256357

Format: PDF / Kindle (mobi) / ePub

Author note: Edited by Michael Stark

Computer graphics is a vast field, and getting larger every day. It is impossible to cover every topic of interest, even within a specialization such as CG rendering. For many years, Noriko Kurachi has reported on the latest developments for Japanese readers in her monthly column for CG World. Being something of a pioneer herself, she selected topics that represented original and promising new directions for research, as opposed to the tried and true methods.

Many of these novel ideas paid off handsomely, and these are the topics covered in this book. Starting from the basic behavior of light, Ms. Kurachi introduces the most useful techniques for global and local illumination using geometric descriptions of an environment in the first section. She then goes on to describe image-based techniques that rely on captured data to do their magic in the second section. In the final section, she looks at the synthesis of these two complementary approaches and what they mean for the future of computer graphics.

"The book you hold today tells the story of this new era of computer graphics. Working closely with researchers who helped lead this revolution, Noriko Kurachi describes these key innovations and brings them together as a coherent body of knowledge. Please read this book, practice the techniques, and figure out if they will allow you to create the visions you have in your mind." -Paul Debevec, pioneer in HDR imaging and image-based modeling





















abruptly. Ray casting can sometimes be optimized by using adaptive sampling (Figure 3.4), where more sample points are taken in sections where the volume data is changing more quickly. The sampling interval is made narrower when the difference in adjacent voxels values is large. Volume Figure 3.4 Adaptive ray sampling works by taking more samples where the density changes more rapidly. 58 3. Volume Rendering and Participating Media The ray-casting technique has become the most common

is enhanced by a good understanding of the materials being visualized. Known conditions such as “fat is never contained in the bone” and “muscle does not exist in the skull” can be applied to better distinguish features when visualizing the human body. Although it was not clearly described in the paper, basing the rendering on the density offers a lot of flexibility in visualization. This is because a lookup table can be added to convert “true” densities of the various materials into virtual

rendering: “The Lumigraph” [Gortler et al. 96] authored by Steven J. Gortler, Radek Grzeszczuk, Richard Szeliski, and Michael F. Cohen, and “Light Field Rendering” by Marc Levoy and Pat Hanrahan [Levoy and Hanrahan 96]. Before these papers were published, IBR techniques used a collection of 2D images to store the acquired photographic images. Renderings from an arbitrary viewpoint were constructed through interpolation or warping of these images. Both the 1996 papers introduced essentially the

framework of the “Fourier Slice Photogra- 162 5. Image-Based Rendering 4D Fourier transform Virtual refocusing in real space (change of basis + integral projection) Virtual refocusing in Fourier space (change of basis + slicing) Inverse 2D Fourier transform Figure 5.26 Diagram of Fourier slice photography. Taking the 4D Fourier transform of the light field, transforming in the Fourier domain (a change of basis), taking a 2D slice, and transforming back is equivalent to a virtual refocus

linear log( X ) X: exposure Figure 6.1 A hypothetical response curve for an electronic sensor. Exposure values in the region labeled “1” are too low to be recorded; values in region “3” cause the senor to saturate and record the maximum value regardless of the exposure. In region “2”, the sensor records a value approximately proportional to the exposure, so it is known as the linear or working range of the device. Region 1, the left part of the sigmoid shape, is typically ignored in

Download sample