I’ve been learning about PBR recently. When using IBL to calculate the diffuse term part, I have some doubts. When calculating the spherical coordinates, I found that most of article are in the z up direction, but in threejs it is in the y up direction. The azimuth is defined as starting from +z. I tried to convolve envMap based on this benchmark, but I got the wrong image. Can someone explain it to me?
result:
Here are the modifications I made：
vec3 sphericalEnvmapToDirection(vec2 tex) {
float theta = PI * (1.0 - tex.t);
float phi = 2.0 * PI * (0.5 - tex.s);
// return vec3(sin(theta) * cos(phi), sin(theta) * sin(phi), cos(theta));
return vec3(sin(theta) * sin(phi), cos(theta),sin(theta) * cos(phi));
}
vec2 directionToSphericalEnvmap(vec3 dir) {
// float phi = atan(dir.y, dir.x);
// float theta = acos(dir.z);
float phi = atan(dir.x, dir.z);
float theta = acos(dir.y);
float s = 0.5 - phi / (2.0 * PI);
float t = 1.0 - theta / PI;
return vec2(s, t);
}
and in for loop
vec3 prefilterEnvMapDiffuse(in sampler2D envmapSampler, in vec2 tex) {
float px = t2p(tex.x, width);
float py = t2p(tex.y, height);
vec3 normal = sphericalEnvmapToDirection(tex);
mat3 normalTransform = getNormalFrame(normal);
vec3 result = vec3(0.0);
uint N = uint(samples);
for(uint n = 0u; n < N; n++) {
vec3 random = random_pcg3d(uvec3(px, py, n));
float phi = 2.0 * PI * random.x;
float theta = asin(sqrt(random.y));
// vec3 posLocal = vec3(sin(theta) * cos(phi), sin(theta) * sin(phi), cos(theta));
vec3 posLocal = vec3(sin(theta) * sin(phi), cos(theta),sin(theta) * cos(phi));
vec3 posWorld = normalTransform * posLocal;
vec2 uv = directionToSphericalEnvmap(posWorld);
vec3 radiance = textureLod(envmapSampler, uv, mipmapLevel).rgb;
result += radiance;
}
result = result / float(N);
return result;
}
This is the axis definition used in most page and literature：
I made an example for anyone to debug：demo
To see the results, click the play button in Time Control.
if u want to modify shader, can do that
select shader node and click edit code button
