In all the cascading shadow map explanations I have read, the final projection into the shadow-map is done using an orthographic projection. This makes sense for a directional light because a directional light "sees" things orthographically.

Spot lights on the other hand "see" things with a perspective transformation. Once we have our light space bounding volume, should we be performing the final projection into the shadow-map using the spotlight's projection matrix, rather than an orthographic projection matrix?

Perhaps I am missing the point here. Cascading shadow maps are necessary to provide shadow maps over large areas, thus because spotlights tend to have a very limited area of effect, perhaps cascading shadow maps are not necessary for spotlights.

  • \$\begingroup\$ I think you hit on it at the end there. We tend to reach for cascaded shadow maps for directional lights because they cover such a large area, with shadow texel density needs highest near the camera and falling off elsewhere. So most implementations you see are solving this problem. That's not to say that you couldn't implement them for spotlights, though the texel density calculation gets a bit more complicated due to the spreading of the light cone itself. Have you observed shadow quality issues with your spotlights that you expect will need cascades to solve? \$\endgroup\$ – DMGregory Aug 27 '18 at 12:56
  • \$\begingroup\$ @DMGregory Thank you for your reply. I have not experienced spotlight quality issues, I was more concerned with the theory. I imagine a CSM would help in cases of having a low angle camera and a low angle spotlight casting onto the ground. Probably not worth the performance hit though. It would be better to redesign your scene. \$\endgroup\$ – Zwander Aug 27 '18 at 23:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.