light direction calibration

[niubiplus2]

Hello, I have read the code, but I had been confused for a long time that the part of the calibration of the light direction return the normal of the ball, not the light direction? Could you explain about it, thanks a lot.

[worker-ljw]

Excuse me，have you solved this problem？ I have the same confusion.

[NewProggie]

The sphere is used to determine the (unknown) light direction assuming Lambert reflection properties. That is, given a pixel intensity (gray value) at position (x,y) and its known surface normal on the sphere geometry, we can solve light directions L as described here: https://en.wikipedia.org/wiki/Lambertian_reflectance

[worker-ljw]

I agree with you, but I found some differences. In this part of the code, I understand ”LightsInv“ as the inverse of the direction of the light source. But when I continued reading the code with this conjecture, I found that it also seems to be the normal direction of the spherical surface.

I can't understand this phenomenon unless the calibration image used in the process of calibrating the direction of the light source requires that the direction of the light source always be towards the center of the sphere.