(new)? lighting theory

[AIRmichael]

Hey, I know instead of calculating the lighting mostly,its done a differant way. The shadow has been calculated. But what about this technique for the lighting:

Create a sphere around the lightsource. ( Afcourse this doesnt have to be rendered ). Increase the radius of the sphere, till every part of the sphere has been clipped at a surface or that r (radius) so big that it doesnt have any effect for the eye. ( Note that useing multiple light sources, it can be seen indeed ).

r = sphere radius = distance from light source to the object.
distance = Xdistance + Ydistance + Zdistance.


The dissapeard area grows with the sphere, so there wont be any clipping "behind" objects. So the clipping area will be x * r + y * r.


//loop through all light sources
for (int l = 0; l < all_light_sources; l++)
{


/*loop through all grid cubes of the sphere (the more detail the sphere has, the more lighting precision)*/

for (int i = 0; i < all_polygons; i++)
{

if (!sphere.Clipped() && 1/(totalR * totalR)> 0.0001f)
// 0.0001f stands for the "viewable" amount of light
{

/* render per pixel , triangle or whatever u want
The sphere has to be a grid. When a "cube" clips with its surface, it does 1/r^2. Colors have to be the same though glLight1[] = {1/r^2,1^r2 etc};*/


}
}
}

Clipped() when the x,y,z coordinates of the world is equal to the sphere coordinates which increases.
So the detail of the lighting in the z depth equals to the growing speed of the sphere. The detail of the lighting in the x,and y axis equals to the polygon amount/detail of the sphere.

For less detail for lighting in the distance, just lower the detail of its sphere. ( can do with 1/r^2 as well ). Wow 1/r^2 can be used for almost anything. Remove the 1 and u got a lightened up a room with black as light source. Same as 1/r^2 -= 2 * (1/r^2); though.
errr lighting = (1/(r^2*r^2)) decreases its lighting effect on a object alot faster. Going to far out of the original subject now.

[Zaldron]

Works for lightmap calculations IMO, but the bubble method is quite slow for real-time applications.

[AIRmichael]

Afcourse its slow, but should be fixed soon enough when new hardware ships.

[Zaldron]

Heh :)
Relying on raw power. I wish that was possible already.

[Doom-Child]

That's an interesting idea...the only thing I see that would improve it speed wise would be to just use three circles. Figuring the outer edge of a circle is much faster than attempting to generate sphere coordinates (something that took me for-freaking-ever to polish), and, provided they were positioned correctly in regards to the eyeplane, would provide almost the same effect. I've always hated lights that shine through walls...

DC