A question has come up recently in Blender SE, remarking on the absence of blur due to depth of field in reflections in a spherical surface. This is a real-life effect.. can anyone come up with a nice explanation of it?
$\begingroup$ Because the virtual image is inside the sphere, not far behind it like for a plane mirror. $\endgroup$– user106Nov 25, 2019 at 17:29
$\begingroup$ @Rahul That's the one, for me.. you should make that an answer? $\endgroup$– Robin BettsNov 25, 2019 at 19:22
There's no visible blur for the same reason you often don't see blur in very wide angle lenses: the circle of confusion is much smaller than a pixel. But it is still there. An extreme example of this is fisheye lenses.
If you had a camera with infinite resolution, such that you could zoom into the reflection on the ball, you would eventually see the blur. And it would be exactly the same blur as if you had aimed a flat mirror at the part of the scene you're looking at through the ball.
Now, we don't have such a camera, but we have something even better: software! So I built this simple setup with two perfectly sharp checkered planes, a sphere and a camera with some depth of field:
Rendering it with some depth of field at 512x512, you'd think the reflection in the sphere is perfectly sharp:
But here is a central crop of the exact same scene rendered at 8192x8192:
And we can see the exact same blur on the checker pattern. It's as simple as that: you don't see the blur because it's too small.
$\begingroup$ Wow.....you could be right... is this true? If you look at the example photo, the direct background seems a comparable size to the reflected background.. but it's hardly a definitive test! I'll try to persuade myself of it by your method before accepting. $\endgroup$ Nov 30, 2019 at 6:53
$\begingroup$ @RobinBetts it isn't the same size. Top: one balcony is about 1/3 the frame width. In the ball: you see a whole building (~6 balconies) in about the same width. Clearly not the same effective field of view. I think I could make a convincing example with a dSLR too. I just need to find a nice shiny ball. $\endgroup$– OlivierNov 30, 2019 at 13:01
$\begingroup$ ..and I've just confirmed your result $\endgroup$ Nov 30, 2019 at 15:20
Because the ball doesn't "see" the world through an imperfect lens system, whereas your camera in that case does. If you were to introduce an imperfect visual system around it, with a resulting blurring point spread function, you will get a blurry reflection.