Aperture of Orthographic camera (orthographic)

[roflmaostc]

Hi!

I noticed that the orthographic camera has no aperture to control the depth of focus whereas the perspective camera has that option by adjusting the aperture size.

Is there any plans to include that? Or is it possible already?

In an 4f system we would need to sample uniformly in an aperture space. The direction of those rays indicates the position of focus:
https://www.mdpi.com/sensors/sensors-20-03116/article_deploy/html/images/sensors-20-03116-g001-550.jpg

I do have some applications in mind where this would be useful ;)

Best,

Felix

[njroussel]

Hi @roflmaostc

No, there is no plan for that. Until you shared that image, I never thought that an orthographic camera model with depth of focus control would actually have any physical interpretation 😅

I don't see any immediate reason why it would be impossible to implement this as a custom sensor.

[roflmaostc]

I think for us it would be really great to have! Orthographic cameras are very standard in machine vision/inspection and standard optics (also some topic we are working on). In fact, our printer system is a 4f system with finite aperture.

I guess, I am lacking a bit the expertise but how much work do you think it would be? Maybe it could be even a semester project? Or too simple?

[njroussel]

I'm assuming these models are well understood and so, given the formulas, it shouldn't take anyone more than a week. I'd say that it's too short for a semester project.

However, if you also want to differentiate the camera parameters, that would be considerably harder.

[roflmaostc]

I think for us we calibrate it before anyway (by measuring the PSF throughout space).

[roflmaostc]

@bathal1 would you be interested in tackling this 👀 ?

[roflmaostc]

An alternative path would be to simply put two lens CAD designs into the optical path before the vials, right?

Does this include a noticeable penalty for doing the ray tracing? I guess no? Since we do later volumetric stepping, the couple of surfaces should not matter a lot?

[njroussel]

It might be noticeable, but it definitely shouldn't be the bottleneck of your pipeline.

I've personally only always used analytical systems for the lenses/sensors. What you suggest can work, I just don't know if you'll run into substantial artifacts from meshing the lens. I guess a fine tessellation should do the trick.