Warp wrote:
> 1) Treat all objects as double-illuminated. The advantage of this is that
>    the decision is very fast to make, and closed surfaces will self-shadow
>    and present no problems. Only open surfaces present problems.
> 
> 2) Calculate whether any part of the area light is visible from the current
>    point by a more complicated algorithm. (Maybe testing against each corner
>    of the area light could be enough, except in the case of the circular
>    area light.) This will make the whole rendering slower because the
>    algorithm would be spawned for each intersection point between a ray
>    and a surface. It also may cause artifacts on the "dark side" of open
>    surfaces, in which case it's not really a solution at all.
> 
> 3) Just treat the area light as if it was a basic grid of point lights
>    which results in perfect lighting (at least if the grid is dense enough),
>    but there would be absolutely no speed benefit in having a specialized
>    area_illumination feature as the exact same effect (including speedwise)
>    could be accomplished by simply creating a grid of point lights.
> 
>   Which one do you prefer?

Well, since 3) would be the same as a grid of area_lights, I consider it 
  no new solution. Seeing how 2) produces artifacts in every case, I 
rule it out. Which leaves me cheering for 1). How often do I really use 
open objects? Rarely, and only under certain circumstances.

The only problem I can see is when you have a semi-transparent object. 
The forced double illumination shows up there. This is no big deal for 
me though.

Thanks~

Sam

Post a reply to this message