Previous | Next --- Slide 44 of 47
Back to Lecture Thumbnails
rbunny

If we keep doing this process, will we end up with an edge detection algorithm? The edges seem to be really bright in the last picture.

rlpo

When dealing with images, we often store image pyramids for downsampled images. Is there an equivalent version of this for meshes?

keenan

@rbunny Yeah, in this particular case you get a bit of a "blurred edge detector" since the filtering is done via cubic interpolation that can over/undershoot the given values. But in general if you used a different filter, you'd get a different result. More to the point: in practice you're probably not repeating the same operations back and forth, but rather applying a sequence of signal processing operations. Enough operations, and you distort the original signal...

keenan

@rlpo Great question! Yes, there's been quite a lot of work on "multiresolution mesh processing," but I wouldn't say there's a single gold standard. As discussed in lecture, geometry is a bigger "zoo" than images, so there are a lot of possible solutions. Here's one classic paper.