An interesting thought about this slide is that a real shadow almost never looks like the leftmost image where its edges are sharp and defined but instead looks like the one on the right where they are a bit blurry. This phenomenon is known as penumbra, which is a display of the wave-like qualities of light. When light is blocked by an physical edge, instead of perfectly coming to a stop, it actually bends and curves around the edge similar to how sound waves don't just disappear around the corner. This is why we see a gradient that extends off the edges of a shadow. Moreover, lights at different wavelengths also bend to different extents. It is interesting that we are modeling this real life phenomenon using random sampling to kinda mimic it instead of modeling the exact physics behind it. Maybe it requires too much computation for a level of realism that is not distinguishable anyways or perhaps we don't even know exactly how the quantum physics behind this works either.
eugene-yoojin-han
I might be missing something, but one thing that comes across my mind is path tracing. Can we just implement global illumination + path tracing with a very high number of bounces to output such a blurry shadow? I feel like the collisions of light rays with the edge of the square will cause some blurred-out edges for the shadow. Please correct me if I'm wrong!
myxamediyar
I think it is cool how all the filters we've discussed before to reduce aliasing could be applied here! I imagine that if we sample enough points using monte carlo and apply a box filter (or some kind of weighted filter based on how close the point is to the middle of the projection of the rectangle), we can basically achieve the true answer. This is very fascinating stuff!
sparky-ed
This is really interesting to see sample random points on light view and the actual answer. Even though the true answer exists, I saw some applications that are using the second example for rendering to give a different experience, and I think it is really cool to see how each stage of making it better to see can be utilized in games' view.
weinatalie
I wonder if there are other ways to achieve the "true answer" shadow on the right apart from sampling more points. For example, could we apply some other anti-aliasing technique we’ve learned previously? Or, instead of sampling 100 points, could we take the shadow that results from sampling 1 point and apply a box blur to approximate a Gaussian blur? Would this pose a greater computational burden than simply sampling more points?
aidangarde
How does randomly sampling a point react with moire patterns? The randomness should eliminate them somewhat, but it still won’t be able to handle the complex patterns at higher resolutions. Is it always better to randomly sample than sample the center? Since it’s not too expensive, it seems like this should always be done
An interesting thought about this slide is that a real shadow almost never looks like the leftmost image where its edges are sharp and defined but instead looks like the one on the right where they are a bit blurry. This phenomenon is known as penumbra, which is a display of the wave-like qualities of light. When light is blocked by an physical edge, instead of perfectly coming to a stop, it actually bends and curves around the edge similar to how sound waves don't just disappear around the corner. This is why we see a gradient that extends off the edges of a shadow. Moreover, lights at different wavelengths also bend to different extents. It is interesting that we are modeling this real life phenomenon using random sampling to kinda mimic it instead of modeling the exact physics behind it. Maybe it requires too much computation for a level of realism that is not distinguishable anyways or perhaps we don't even know exactly how the quantum physics behind this works either.
I might be missing something, but one thing that comes across my mind is path tracing. Can we just implement global illumination + path tracing with a very high number of bounces to output such a blurry shadow? I feel like the collisions of light rays with the edge of the square will cause some blurred-out edges for the shadow. Please correct me if I'm wrong!
I think it is cool how all the filters we've discussed before to reduce aliasing could be applied here! I imagine that if we sample enough points using monte carlo and apply a box filter (or some kind of weighted filter based on how close the point is to the middle of the projection of the rectangle), we can basically achieve the true answer. This is very fascinating stuff!
This is really interesting to see sample random points on light view and the actual answer. Even though the true answer exists, I saw some applications that are using the second example for rendering to give a different experience, and I think it is really cool to see how each stage of making it better to see can be utilized in games' view.
I wonder if there are other ways to achieve the "true answer" shadow on the right apart from sampling more points. For example, could we apply some other anti-aliasing technique we’ve learned previously? Or, instead of sampling 100 points, could we take the shadow that results from sampling 1 point and apply a box blur to approximate a Gaussian blur? Would this pose a greater computational burden than simply sampling more points?
How does randomly sampling a point react with moire patterns? The randomness should eliminate them somewhat, but it still won’t be able to handle the complex patterns at higher resolutions. Is it always better to randomly sample than sample the center? Since it’s not too expensive, it seems like this should always be done