Foreword
Ten years ago there were two worlds of research that rarely crossed. My world, the
world of computer graphics, tried to solve a well-defined problem. Given the geo-
metrical and material description of a virtual scene, a definition of the light sources
in that scene, and a virtual camera, create an image that looks as close as possible to
one that a real camera in a real version of the described scene would look like. In the
other research world, that of computer vision, researchers were struggling with the
opposite question. Given one or more real images of a real scene, get the computer
to describe the geometrical, material, and lighting properties in that real scene.
The computer graphics problem was having great success, to a point. Surprising
to many of us, the real problem derived from a realization that the geometric com-
plexity of the real world overwhelmed our abilities to create it in current geometric
modelers. Consider a child's fuzzy stuffed lion. To create an accurate image of such
a lion could require the daunting task of describing every strand of hair and how it
reflects and transmits light. This is about when we went to knock on the doors of
our computer vision colleagues. We hoped, a bit prematurely, that perhaps we could
just point cameras at the fuzzy lion and a computer vision algorithm could spit out
the full geometric description of the lion. The state-of-the-art was (and still is) that
computer vision algorithms could give us an approximate model of the scene but not
the kind of detail we needed. In some ways we were lucky, for if they had given
us the full model in all its detail, the computer graphics rendering algorithms most
likely could not deal with such complex models.
It was this point that the ideas underlying image based rendering were born. The
idea is that by using the partial success of computer vision algorithms PLUS keeping
the original pixels contained in the input images, one could then leverage the partial
success of computer graphics. I feel quite honored to have been a part of making
some of the first realistic synthetic images of a child's stuffed lion in just this way.
The results that have grown from this initial idea have been quite astounding.
The book you are about to read introduces the reader to the rich collaboration
that has taken place over the past decade at the intersection of computer graphics
and computer vision. The authors have lived in this exciting research world and have
produced many of
the
seminal papers in the field. The book provides both a historical
评论0