Trapezoidal Shadow Maps?

Nick[FM]

Regular
Just found this, and thought you guys might want to see it too. Not sure if it has been discussed before? If it has, sorry! :)

http://www.comp.nus.edu.sg/~tants/tsm.html

With TSM, the animation is smooth with only occasional shadow flickering for both static (see Figure d(i)) and dynamic objects (see Figure d(ii) and d(iii)) and has high quality shadows for all large (see Figure d(ii)) as well as small objects (see Figure d(iv)), curved (see the mushrooms in Figure d(i) and Figure d(iv)) and sharp objects (see the lance in Figure d(ii) and the axe in Figure d(iii)), and for objects from near to far within an eye's view (see Figure d(i)). In the mpeg videos, the polygon offset bias used by TSM is visible as a gap between the leg of the pergola and the terrain. This is of the same amount as that of SSM, whereas the bias needed for BB and PSM are much larger and thus a larger gap between the leg of the pergola and the terrain. Though at some specific shots, the quality of the TSM shadows may not be much superior than that of BB or PSM, the experience of the flythrough is very pleasant—shadows are now natural parts of the scene that no longer attract special (undesirable) attention!

Has anyone more info about this technique, or running demos?
 
GameCat said:
http://www.gamedev.net/community/forums/topic.asp?topic_id=179941
Thanks for the link, but it seems that they discuss more PSM and not that much TSM. Anyway, a good read. Maybe Humus could make us a running demo with TSM? ;) It'd be nice to see it running as demo instead of videos and screenshots.
 
There's a lot of discussion in that thread on various ways to utilize shadow map precision better. A lot of has to do with distorting trapezoids into rectangles. There is also some discussion about what trapezoid shadow maps actually do. So nothing that directly explains what trapezoidal shadow maps actually are, but lots of stuff on how they might work and lots of tips for getting better shadow map precision.
 
That's true. I just think that the ways to improve the quality of PSM (using the 5 shadow map idea) isn't very cost effective, perfomance-wise. From what I have understood (which isn't all that much) the trapezoidal shadow maps should be better in every way compared to trditional PSM. Even getting PSM to work flawlessly (at least almost) seem to be very difficult.
 
Interesting, I'm also interested in any paper/book/article/demo on the subject.
 
Back
Top