Not completely relevant, but interestingly, this model reminds me of an ancient theory of how humans perceive the world we live in called the Emission Theory (https://en.wikipedia.org/wiki/Emission_theory_(vision)); the theory assumes human eyes emit rays to perceive objects. Of course, I understand that in this slide, what we are assuming is a totally different thing, since we can actually imagine the image plane and the objects on the same side of the focal point for a better illustration.
Carpetfizz
If we wanted to simulate viewing the scene through a lens, would the ray tracing algorithm involve projecting rays through through some kind of surface defined by the properties of the lens and then "bending" the outgoing ray according to Snell's law?
What would be some ways we can define the curvature of the lens mathematically?
If this technique works it can also be extended to support the simulation of multiple lenses.
StaffJakeHoles
Yes this is definitely possible. In fact you will include a lens in proj 3-2 to simulate depth of field! In our case, we assume that the lens is ideal so we can calculate its effect using the thin lens equation. In reality every lens has at least some aberration so this approximation is not completely honest. In some applications it is enough but for say a camera manufacturer, a more realistic model would be necessarily. I can't speak on how you would model realistic aberration though.
DavidVakshlyak
For the project, once we get a coordinate in in [0, 1]^2 sensor space, how would we generate the ray? I think we would use the camera position as the origin and difference between that sensor coordinate and the camera position as the direction. Normalized of course.
Not completely relevant, but interestingly, this model reminds me of an ancient theory of how humans perceive the world we live in called the Emission Theory (https://en.wikipedia.org/wiki/Emission_theory_(vision)); the theory assumes human eyes emit rays to perceive objects. Of course, I understand that in this slide, what we are assuming is a totally different thing, since we can actually imagine the image plane and the objects on the same side of the focal point for a better illustration.
If we wanted to simulate viewing the scene through a lens, would the ray tracing algorithm involve projecting rays through through some kind of surface defined by the properties of the lens and then "bending" the outgoing ray according to Snell's law?
What would be some ways we can define the curvature of the lens mathematically?
If this technique works it can also be extended to support the simulation of multiple lenses.
Yes this is definitely possible. In fact you will include a lens in proj 3-2 to simulate depth of field! In our case, we assume that the lens is ideal so we can calculate its effect using the thin lens equation. In reality every lens has at least some aberration so this approximation is not completely honest. In some applications it is enough but for say a camera manufacturer, a more realistic model would be necessarily. I can't speak on how you would model realistic aberration though.
For the project, once we get a coordinate in in [0, 1]^2 sensor space, how would we generate the ray? I think we would use the camera position as the origin and difference between that sensor coordinate and the camera position as the direction. Normalized of course.