The even/odd intersection count test mentioned in lecture disregards the possibility of a ray being tangent to a surface, right?
justaddwater
I think that the intersection test determines whether a point is inside or outside of the object. So with a ray that is tangent to the surface of the object, there would be 1 point of intersection. So I guess according to this definition, a point on the surface of the object is seen as a point 'inside' the object, as opposed to being 'outside' (so not on the surface nor on the inside of the object)? If so, then I think this count test still works.
rilakkuma
Still confused about the inside-outside test, what is the orientation and the starting point of the ray when doing the test?
JimL
@justaddwater What if the geometry is concave and there are two intersections and one of them is tangent to the surface? But still I think if we are solving the quadratic equation or higher order equation to get the intersect points, we can tell whether the point is tangent to the surface or not.
Max
The ray becoming tangent to the surface isn't considered "piercing" here, so we would want to disregard any tangential solutions.
@rilakkuma those are both arbitrary - any point inside the surface will produce an odd number of intersections in any direction, and any point outside the surface will produce an even number. (Assuming the surface is closed and orientable, of course - no Klein bottles here.)
allai5
How does ray-tracing account for the brightness of some point on the surface?
wenere
Tangent is really a tricky position... Is this kind of position causing some bugs in existing rendering engine?
Fjorge
Does a ray "bounce" upon hitting a surface? If so, how would you calculate the loss of brightness of the ray after a number of bounces?
jasonx
Something to do with the fact that "inside" cannot extend infinitely while outside does. So we count transitions between inside and outside.
Max
@allai5 @Fjorge you can compute this by sampling the surface's BRDF, which defines how light interacts with the material (lecture coming soon).
The even/odd intersection count test mentioned in lecture disregards the possibility of a ray being tangent to a surface, right?
I think that the intersection test determines whether a point is inside or outside of the object. So with a ray that is tangent to the surface of the object, there would be 1 point of intersection. So I guess according to this definition, a point on the surface of the object is seen as a point 'inside' the object, as opposed to being 'outside' (so not on the surface nor on the inside of the object)? If so, then I think this count test still works.
Still confused about the inside-outside test, what is the orientation and the starting point of the ray when doing the test?
@justaddwater What if the geometry is concave and there are two intersections and one of them is tangent to the surface? But still I think if we are solving the quadratic equation or higher order equation to get the intersect points, we can tell whether the point is tangent to the surface or not.
The ray becoming tangent to the surface isn't considered "piercing" here, so we would want to disregard any tangential solutions.
@rilakkuma those are both arbitrary - any point inside the surface will produce an odd number of intersections in any direction, and any point outside the surface will produce an even number. (Assuming the surface is closed and orientable, of course - no Klein bottles here.)
How does ray-tracing account for the brightness of some point on the surface?
Tangent is really a tricky position... Is this kind of position causing some bugs in existing rendering engine?
Does a ray "bounce" upon hitting a surface? If so, how would you calculate the loss of brightness of the ray after a number of bounces?
Something to do with the fact that "inside" cannot extend infinitely while outside does. So we count transitions between inside and outside.
@allai5 @Fjorge you can compute this by sampling the surface's BRDF, which defines how light interacts with the material (lecture coming soon).