home *** CD-ROM | disk | FTP | other *** search
- In article <D9rKrn.5Cy@cix.compulink.co.uk>,
- Glyn Williams <glynw@cix.compulink.co.uk> wrote:
- >John,
- > (I am a programmer in my other life)
- > As far as I understand it, Lightwave's adaptive sampling works by
- >examining the first (aliased) image generated and then fixes pixels which
- >it decides might be candidates for anti-aliasing.I think it does this by
- >looking for "edge" pixels which neigbor pixels of different intensities.
- >My concern was based on these circumstances:
- > An aliasing error which mistakenly gives rise to flat colour
- >which adaptive sampling would then ignore.
- > For instance consider a simple grating of regular white polygons.
- >The first pass might render the entire grating as a solid white area -
- >this would not be picked up.
- > Like you say, this can always be turned off.
- >
- > This wouldn't be a problem with 3DS - but then again, there are
- >*lots* of other things wrong with 3DS.
-
- > Glyn Williams
-
- Unless I miss my guess, by a "grating" you mean a series of white polygons
- arranged in a grid of parallel lines? (roughly like an American football
- gridiron?) If that is what you mean, I would suggest that that is not the
- case.
-
- LW's anti-aliasing method DOES NOT involve the image itself except for the
- "searching for edges" stage. If you think of each pixel as a line radiating
- out from the camera point into space, anti-aliasing actually renders
- additional lines around the originals in order to calculate the proper
- amounts of anti-aliasing to use.
-
- Now, consider the aliasing effect caused by a sample rate which is too close
- to natural frequencies present in the image. Common examples of this are
- moire patterns in print, and that wierd "electric" look seen in video when the
- subject has lines in it which are very close together (such as distant
- buildings, zoot suits or hocky referee's jerseys). In fact, the term
- "aliasing" was originally coined to describe these wierd effects which are
- basically spurious frequencies introduced by the sampling process. The
- "jaggies" most commonly associated with the term "anti-aliasing" in video
- and rendering are in fact different artifacts of sampling; they are
- quantization effects, which is more a factor of resolution and can occur
- without the kind of false signals caused by bad sampling.
-
- LightWave's use of this more fundamental technique, although MUCH more
- process-intensive than mere jaggie-reduction, has a PHENOMENALLY more
- useful effect IMO. In a project here at Vivid, I rendered a cityscape
- with distant buildings with simple horizontal lines of windows (some lit,
- some not). The game required the final product to be in 320x200 resolution.
-
- Renders with no anti-aliasing produced these wierd, gigantic candy stripes
- all over my buildings! This was due to the fact that the visual size of
- the lines of windows were very close to that of the pixel lines, i.e. the
- frequency of the windows was very close to that of the sample size (render
- resolution).
-
- Two-dimensional anti-aliasing on that image would result in the huge solid
- areas as you describe (with smooth edges :^). But, when I render that image
- using AA set to Medium in LW, the result is remarkably decent, in that the
- solid areas are mostly gone (they can still be seen, though.) When viewed
- from a sufficient distance, the anti-aliased image is a much closer
- approximation to a higher-resolution equivalent viewed at the same distance
- than would the image-based result.
-
- Rendering to twice the resolution and then scaling down in ADPro is
- functionally the same thing, except that the Adaptive Sampling option helps
- eliminate the redundant parts from rendering time.
-
- I am curious to know whether anti-aliasing in 3DS causes the same kind of
- increases in rendering time; I have sen anti-aliased images produced in
- 3D Studio here, and its AA looked impressive to me.
-
- Jim May
- The Vivid Group
- Mandala VR Systems
-
-
-
-
-