Reflective instead of transmissive?

[wjhrdy]

I'm interested in modifying this software to make a reflective object instead of a transmissive object. I'm wondering if there are fundamental limitations in this project that would limit my ability to do that.

Thanks!

[dylanmsu]

Hi,

It is certainly possible to modify it for reflective caustics. I have implemented the math for it but you have to change the setup like sunlight direction and maybe even focal plane offset.

I haven't had the time yet to make a proper abstraction and interface for it. Also when the focal plane is not directly in front of the reflective surface you can get issues with distortion in the reflected image.

To enable reflective caustics you have to set the reflective caustics flag to true in fresnelMapping() in apps/caustic_design.cpp, and flip the sign of the light direction eg incidentLight[2] = 1.0. If the solver fails to give correct results, you can try the negative sign again.

Kind regards

[Graysonnwu]

Also when the focal plane is not directly in front of the reflective surface you can get issues with distortion in the reflected image.

@dylanmsu Exactly! And this isn't just a problem with reflective caustics. As long as there is some rotation between the target plane and source plane, you'll get this kind of distortion.

On the other hand, simply moving or scaling the target plane won't introduce distortion. I suspect this is also due to curl. Translation and scaling don't introduce curl, but rotation around any axis will.

I have no idea how to solve this issue though. I noticed you updated the README today with 3 potential tranport solving strategies. I wonder whether any of those might help?

[dylanmsu]

Hi @Graysonnwu, i included the description of potential solvers to hopefully protect against patents. I don't think these will fix that specific distortion issue. You will probably need to do full 3d optimal transport like 3d power diagrams, or use a method with differentiable rendering.