Next: 4.3.3 Linear Interval Arithmetic
Up: 4.3.1 Optimization: Super-Pixel Rendering
Previous: 4.3.1 Optimization: Super-Pixel Rendering
The super-pixel rendering algorithm maintains
a list L, of pixel clusters.
Initially, L consists of a single cluster of pixels;
that cluster describes all of R.
The rendering starts with all pixels set to .
On each iteration, a cluster is removed from L;
is then evaluated:
- If
then is set to for all .
- If
then is set to for all .
- If
then is cut into several subclusters
with . All of the new clusters
are added to L. The cuts are performed along pixel boundaries
so that all pixels belong to at most one member of L.
Subdivision is not performed on clusters which describe single
pixels; the pixel testing methods outlined earlier are performed
on single pixel clusters.
An example rendering, produced using super-pixel testing,
follows:
denotes a super-pixel rendering,
where L consists of clusters of pixels.
Section will provide a motivation for keeping
cluster cuts nicely aligned, as was done above.
All intermediate renderings are renderings
of ,
and may be presented to the user.
Another example super-pixel rendering follows:
Next: 4.3.3 Linear Interval Arithmetic
Up: 4.3.1 Optimization: Super-Pixel Rendering
Previous: 4.3.1 Optimization: Super-Pixel Rendering