TEST(BSDFSampling, TR_std_0p2_0p1) {
TestBSDF([](BSDF* bsdf, MemoryArena& arena) -> void {
createMicrofacet(bsdf, arena, false, false, 0.2, 0.1);
}, "Trowbridge-Reitz, std sample, alpha = 0.2/0.1");
}
TEST(BSDFSampling, Beckmann_std_0p2_0p1) {
TestBSDF([](BSDF* bsdf, MemoryArena& arena) -> void {
createMicrofacet(bsdf, arena, true, false, 0.2, 0.1);
}, "Beckmann, std sample, alpha = 0.2/0.1");
}
TEST(BSDFSampling, Beckmann_VA_0p5) {
TestBSDF([](BSDF* bsdf, MemoryArena& arena) -> void {
createMicrofacet(bsdf, arena, true, true, 0.5, 0.5);
}, "Beckmann, visible area sample, alpha = 0.5");
}
TEST(BSDFSampling, TR_VA_0p3_0p15) {
TestBSDF([](BSDF* bsdf, MemoryArena& arena) -> void {
createMicrofacet(bsdf, arena, false, true, 0.3, 0.15);
}, "Trowbridge-Reitz, visible area sample, alpha = 0.3/0.15");
}
int main(int argc, char* argv[]) {
Options opt;
pbrtInit(opt);
// number of monte carlo estimates
// const int estimates = 1;
const int estimates = 10000000;
// radiance of uniform environment map
const double environmentRadiance = 1.0;
fprintf(stderr,
"outgoing radiance from a surface viewed\n"
"straight on with uniform lighting\n\n"
" uniform incoming radiance = %.3f\n"
" monte carlo samples = %d\n\n\n",
environmentRadiance, estimates);
CreateBSDFFunc BSDFFuncArray[] = {
createLambertian,
createOrenNayar0,
createOrenNayar20,
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, true, 0.5, 0.5); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, true, 0.5, 0.5); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, true, 0.2, 0.1); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, true, 0.2, 0.1); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, true, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, true, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, true, 0.33, 0.033); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, true, 0.33, 0.033); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, false, 0.5, 0.5); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, false, 0.5, 0.5); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, false, 0.2, 0.1); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, false, 0.2, 0.1); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, false, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, false, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, true, false, 0.33, 0.033); },
[](BSDF* bsdf) -> void
{ createMicrofacet(bsdf, false, false, 0.33, 0.033); },
[](BSDF* bsdf) -> void
{ createFresnelBlend(bsdf, true, true, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createFresnelBlend(bsdf, false, true, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createFresnelBlend(bsdf, true, false, 0.15, 0.25); },
[](BSDF* bsdf) -> void
{ createFresnelBlend(bsdf, false, false, 0.15, 0.25); },
};
const char* BSDFFuncDescripArray[] = {
"Lambertian",
"Oren Nayar (sigma 0)",
"Oren Nayar (sigma 20)",
"Beckmann (roughness 0.5, sample visible mf area)",
"Trowbridge-Reitz (roughness 0.5, sample visible mf area)",
"Beckmann (roughness 0.2/0.1, sample visible mf area)",
"Trowbridge-Reitz (roughness 0.2/0.1, sample visible mf area)",
"Beckmann (roughness 0.15/0.25, sample visible mf area)",
"Trowbridge-Reitz (roughness 0.15/0.25, sample visible mf area)",
"Beckmann (roughness 0.33/0.033, sample visible mf area)",
"Trowbridge-Reitz (roughness 0.33/0.033, sample visible mf area)",
"Beckmann (roughness 0.5, traditional sample wh)",
"Trowbridge-Reitz (roughness 0.5, traditional sample wh)",
"Beckmann (roughness 0.2/0.1, traditional sample wh)",
"Trowbridge-Reitz (roughness 0.2/0.1, traditional sample wh)",
"Beckmann (roughness 0.15/0.25, traditional sample wh)",
"Trowbridge-Reitz (roughness 0.15/0.25, traditional sample wh)",
"Beckmann (roughness 0.33/0.033, traditional sample wh)",
"Trowbridge-Reitz (roughness 0.33/0.033, traditional sample wh)",
"Fresnel Blend Beckmann (roughness 0.15/0.25, sample visible mf area)",
"Fresnel Blend Trowbridge-Reitz (roughness 0.15/0.25, sample visible mf area)",
"Fresnel Blend Beckmann (roughness 0.15/0.25, traditional sample wh)",
"Fresnel Blend Trowbridge-Reitz (roughness 0.15/0.25, traditional sample wh)",
};
GenSampleFunc SampleFuncArray[] = {
Gen_Sample_f,
// CO Gen_CosHemisphere,
// CO Gen_UniformHemisphere,
};
const char* SampleFuncDescripArray[] = {
"BSDF Importance Sampling",
// CO "Cos Hemisphere",
// CO "Uniform Hemisphere",
};
int numModels = sizeof(BSDFFuncArray) / sizeof(BSDFFuncArray[0]);
int numModelsDescrip =
sizeof(BSDFFuncDescripArray) / sizeof(BSDFFuncDescripArray[0]);
int numGenerators = sizeof(SampleFuncArray) / sizeof(SampleFuncArray[0]);
int numGeneratorsDescrip =
sizeof(SampleFuncDescripArray) / sizeof(SampleFuncDescripArray[0]);
if (numModels != numModelsDescrip) {
fprintf(stderr,
"BSDFFuncArray and BSDFFuncDescripArray out of sync!\n");
exit(1);
}
if (numGenerators != numGeneratorsDescrip) {
fprintf(stderr,
"SampleFuncArray and SampleFuncDescripArray out of sync!\n");
exit(1);
}
// for each bsdf model
for (int model = 0; model < numModels; model++) {
BSDF* bsdf;
// create BSDF which requires creating a Shape, casting a Ray
// that hits the shape to get a SurfaceInteraction object.
{
Transform t = RotateX(-90);
bool reverseOrientation = false;
ParamSet p;
std::shared_ptr<Shape> disk(
new Disk(new Transform(t), new Transform(Inverse(t)),
reverseOrientation, 0., 1., 0, 360.));
Point3f origin(
0.1, 1, 0); // offset slightly so we don't hit center of disk
Vector3f direction(0, -1, 0);
Float tHit;
Ray r(origin, direction);
SurfaceInteraction isect;
disk->Intersect(r, &tHit, &isect);
bsdf = ARENA_ALLOC(arena, BSDF)(isect);
(BSDFFuncArray[model])(bsdf);
}
// facing directly at normal
Vector3f woL = Normalize(Vector3f(0, 0, 1));
Vector3f wo = bsdf->LocalToWorld(woL);
// was bsdf->dgShading.nn
const Normal3f n = Normal3f(bsdf->LocalToWorld(Vector3f(0, 0, 1)));
// for each method of generating samples over the hemisphere
for (int gen = 0; gen < numGenerators; gen++) {
double redSum = 0.0;
const int numHistoBins = 10;
double histogram[numHistoBins][numHistoBins];
for (int i = 0; i < numHistoBins; i++) {
for (int j = 0; j < numHistoBins; j++) {
histogram[i][j] = 0;
}
}
int badSamples = 0;
int outsideSamples = 0;
int warningTarget = 1;
for (int sample = 0; sample < estimates; sample++) {
Vector3f wi;
Float pdf;
Spectrum f;
// sample hemisphere around bsdf, wo is fixed
(SampleFuncArray[gen])(bsdf, wo, &wi, &pdf, &f);
double redF = spectrumRedValue(f);
// add hemisphere sample to histogram
Vector3f wiL = bsdf->WorldToLocal(wi);
float x = Clamp(wiL.x, -1.f, 1.f);
float y = Clamp(wiL.y, -1.f, 1.f);
float wiPhi = (atan2(y, x) + Pi) / (2.0 * Pi);
float wiCosTheta = wiL.z;
bool validSample = (wiCosTheta > 1e-7);
if (wiPhi < -0.0001 || wiPhi > 1.0001 || wiCosTheta > 1.0001) {
// wiCosTheta can be less than 0
fprintf(stderr, "bad wi! %.3f %.3f %.3f, (%.3f %.3f)\n",
wiL[0], wiL[1], wiL[2], wiPhi, wiCosTheta);
} else if (validSample) {
int histoPhi = (int)(wiPhi * numHistoBins);
if (histoPhi == numHistoBins)
--histoPhi;
int histoCosTheta = (int)(wiCosTheta * numHistoBins);
if (histoCosTheta == numHistoBins)
--histoCosTheta;
assert(histoPhi >= 0 && histoPhi < numHistoBins);
assert(histoCosTheta >= 0 && histoCosTheta < numHistoBins);
histogram[histoCosTheta][histoPhi] += 1.0 / pdf;
}
if (!validSample) {
outsideSamples++;
} else if (pdf == 0.f || std::isnan(pdf) || redF < 0 ||
std::isnan(redF)) {
if (badSamples == warningTarget) {
fprintf(stderr,
"warning %d, bad sample %d! "
"pdf: %.3f, redF: %.3f\n",
warningTarget, sample, pdf, redF);
warningTarget *= 10;
}
badSamples++;
} else {
// outgoing radiance estimate =
// bsdf * incomingRadiance * cos(wi) / pdf
redSum += redF * environmentRadiance * AbsDot(wi, n) / pdf;
}
}
int goodSamples = estimates - badSamples;
// print results
fprintf(stderr,
"*** BRDF: '%s', Samples: '%s'\n\n"
"wi histogram showing the relative weight in each bin\n"
" all entries should be close to 2pi = %.5f:\n"
" (%d bad samples, %d outside samples)\n\n"
" phi bins\n",
BSDFFuncDescripArray[model], SampleFuncDescripArray[gen],
Pi * 2.0, badSamples, outsideSamples);
double totalSum = 0.0;
for (int i = 0; i < numHistoBins; i++) {
fprintf(stderr, " cos(theta) bin %02d:", i);
for (int j = 0; j < numHistoBins; j++) {
fprintf(stderr, " %5.2f", histogram[i][j] * numHistoBins *
numHistoBins / goodSamples);
totalSum += histogram[i][j];
}
fprintf(stderr, "\n");
}
fprintf(stderr,
"\n final average : %.5f (error %.5f)\n\n"
" radiance = %.5f\n\n",
totalSum / goodSamples, totalSum / goodSamples - Pi * 2.0,
redSum / goodSamples);
}
}
pbrtCleanup();
return 0;
}
