I'm looking at the rendering equation and the source code for a classical ray tracer (so no Monte Carlo or anything yet included), included below in short pseudo code for clarity.
I have trouble understanding why the cosine from the rendering equation isn't included in the ray tracer implementation. I have included a comment on where I think it's missing.
I can see that with the BRDF for perfect specular reflection (f_r = (δ(cosθi − cosθr)δ(φi − φr ± π)) / cosθi) the cosine gets cancelled out. But what about it missing in combination with the phong reflection model as shown below?
for (loop through pixels)
{
Ray ray(eye, getDirection(col+0.5,row+0.5));
vec3 color = trace(ray, 0);
}
trace(ray, recursion_depth)
{
if (too deep, or ray hits nothing)
return vec3(0,0,0); // return black
Hit hit(ray);
return vec3(hit.Ed) // light source color
+ classic_trace_lights(hit) // direct light
+ classic_trace_reflection(hit, depth) // specular reflection
+ classic_trace_refraction(hit, depth); // specular refraction
}
classic_trace_reflection(hit, depth)
{
....
vec3 R = ideal_reflected_direction(hit.vdir, hit.normal);
Ray ray(hit.pos, R);
return hit.Rs * classic_trace_recursive(ray, depth+1); // Rs = specular color
}
classic_trace_lights(...) { ... }
classic_trace_light(...)
{
....
return phong(...) * light.Ed /* * COSINE MISSING HERE? */;
}
....