Part IV
Systems and Applications
In the last part of the book, we detail four systems that acquire and render scenes
using different types of representations. The representations featured in this part are
geometryless (Chapters 14 and 17), with geometric proxies (Chapter 15), and layers
(Chapter 16).
Chapter 14 (Rendering by Manifold Hopping) shows how manifold mosaics or
multiperspective panoramas can be made more compact. This is accomplished by us-
ing the simple observation that humans perceive motion as continuous if the change
in scene is small enough. By capitalizing on this observation, the minimal number of
sampled manifolds can be derived; the scene is then rendered by "hopping" across
manifolds. No explicit geometry is used.
Constructing an image-based representation of large environments would, in
principle, require a massive amount of image data to be captured. However, we can
usually make the reasonable assumption that some parts of the environment are more
interesting than others. Areas of higher interest would then be captured using more
images and afforded higher degrees of freedom in (local) navigation. Chapter 15
(Large Environment Rendering using Plenoptic Primitives) describes an authoring
and rendering system that generates a combination of panoramic videos and Con-
centric Mosaics (CMs). CMs are placed in areas of more significant interest, and
they are connected by panoramic videos.
Chapter 16 (Pop-Up Light Field: An Interactive Image-Based Modeling and Ren-
dering System) shows how a layered-based representation can be extracted from a
series of images. The key is to allow the user indicate areas where artifacts occur.
Alpha matting at the layer boundaries and highlighted areas are then estimated to en-
sure coherence across the images and artifact-free rendering. Hardware-accelerated
rendering ensures that the editing process is truly interactive.
One characteristic of
a
light field is that it is difficult to edit and manipulate. This
limits the appeal of light fields. Chapter 17 (Feature-Based Light Field Morphing)
describes an interactive system that allows users to morph one light field into another
using a series of simple feature associations. This technique preserves the capability
剩余79页未读，继续阅读
评论0