Parti
Representations and Rendering Techniques
The first part of the book is a broad survey of IBR representations and rendering tech-
niques. While there is significant overlap between the type of representation and the
rendering mechanism, we chose to highlight representational and rendering issues in
separate chapters. We devote two chapters to representations: one for (mostly) static
scenes, and another for dynamic scenes. (Other relevant surveys on IBR can be found
in [155, 339, 345].)
Unsurprisingly, the earliest work on IBR focused on static scenes, mostly due
to hardware limitations in image capture and storage. Chapter 2 describes IBR rep-
resentations for static scenes. More importantly, it sets the stage for other chapters
by describing fundamental issues such as the plenoptic function and how the repre-
sentations are related to it, classifications of representations (no geometry, implicit
geometry, explicit geometry), and the importance of view-dependency.
Chapter 3 follows up with descriptions of systems for rendering dynamic scenes.
Such systems are possible with recent advancements in image acquisition hardware,
higher capacity drives, and faster PCs. Virtually all these systems rely on extracted
geometry for rendering due to the limit in the number of cameras. It is interesting
to note their different design decisions, such as generating global 3D models on a
per-timeframe basis versus view-dependent layered geometries, and freeform shapes
versus model-based ones. The different design decisions result in varying rendering
complexity and quality.
The type of rendering depends on the type of representation. In Chapter 4, we
partition the type of rendering into point-based, layer-based, and monolithic-based
rendering. (By monolithic, we mean single geometries such as 3D meshes.) We
describe well-known concepts such as forward and backward mapping and ray-
selection strategies. We also discuss hardware rendering issues in this chapter.
剩余82页未读，继续阅读
评论1