Next: Conclusions
Up: Multilayer Images for Interactive
Previous: Sample application: Anatomy Browser
Various extensions of the basic MLI format can be made to provide
additional functionality, at the expense of slightly increased
transmission and rendering costs:
- The Z value could be stored at each pixel (in addition to the
intensity value) to provide 3D coordinates of pick points.
- If interaction with the lighting and/or more realistic specular
coloring is desired, the intensity value for each pixel could instead
be made to encode the surface normal. Material properties and light
source positions could be used to compute for each structure a color
lookup table with entries for each possible surface normal. By moving
lighting computations to the client, we increase the rendering time
but add some interactive functionality and realism.
- To eliminate the problem of invisible backfaces, a separate
rendering of each structure with front-face culling could be recorded,
and incorporated in the same way as any other structure. In
semi-transparent renderings, the combination of these front and back
shells would represent the whole structure more accurately.
Finally, it might be worth investigating the use of MLIs in
combination with existing image-based techniques to permit more types
of interaction than either method supports alone. If an MLI was
substituted for each frame required by an existing image-based
algorithm, then structure selection, transparency, and coloring could
be accomplished in a pre-processing step by our rendering algorithm;
image warping and interpolation could then be handled as before. This
type of system could also be seen as an extension of the work in
[9].
Chris Umans
Sun Sep 7 15:06:59 PDT 1997