Because of aliasing effects, you could get random shadows in pits, right? (E.g. z = sin(x)sin(y)). If your shadow map only happens to sample peaks, and the target point is not on a peak, then you would see everything in shadow using this algorithm.
ecohen2
Is there any chance you can explain the 2nd and third steps more? I just tried rewatching this part in lecture and still don't quite get how (or why it is necessary) to transform that P into a different p for the shadow map. I.e. how do we know that those are the transformations desired?
More generally, I feel like I am a little confused with the phrase "P into Pproj" as don't we need to define the projection?
isalinas
@ecohen2 As I understand it, we do one pass from the light source's perspective, sample the distances from the light source, and store this as a texture map. It's necessary to transform P into P_proj in order to determine where it lies in this texture map. Once we know where P falls in the light source's texture map (coordinates given by P_proj), we interpolate its grayscale value to eliminate aliasing and this tells us how much light P is receiving. This gives an idea as to why it's necessary to do this.
I can't answer the "how" question clearly, as I'm learning this all myself, but you may want to check out the texture slides and some of the following lectures to gain an intuition for how this might occur.
Because of aliasing effects, you could get random shadows in pits, right? (E.g. z = sin(x)sin(y)). If your shadow map only happens to sample peaks, and the target point is not on a peak, then you would see everything in shadow using this algorithm.
Is there any chance you can explain the 2nd and third steps more? I just tried rewatching this part in lecture and still don't quite get how (or why it is necessary) to transform that P into a different p for the shadow map. I.e. how do we know that those are the transformations desired?
More generally, I feel like I am a little confused with the phrase "P into Pproj" as don't we need to define the projection?
@ecohen2 As I understand it, we do one pass from the light source's perspective, sample the distances from the light source, and store this as a texture map. It's necessary to transform P into P_proj in order to determine where it lies in this texture map. Once we know where P falls in the light source's texture map (coordinates given by P_proj), we interpolate its grayscale value to eliminate aliasing and this tells us how much light P is receiving. This gives an idea as to why it's necessary to do this. I can't answer the "how" question clearly, as I'm learning this all myself, but you may want to check out the texture slides and some of the following lectures to gain an intuition for how this might occur.