Cycles: Implement denoising option for reducing noise in the rendered image

This commit contains the first part of the new Cycles denoising option,
which filters the resulting image using information gathered during rendering
to get rid of noise while preserving visual features as well as possible.

To use the option, enable it in the render layer options. The default settings
fit a wide range of scenes, but the user can tweak individual settings to
control the tradeoff between a noise-free image, image details, and calculation
time.

Note that the denoiser may still change in the future and that some features
are not implemented yet. The most important missing feature is animation
denoising, which uses information from multiple frames at once to produce a
flicker-free and smoother result. These features will be added in the future.

Finally, thanks to all the people who supported this project:

- Google (through the GSoC) and Theory Studios for sponsoring the development
- The authors of the papers I used for implementing the denoiser (more details
  on them will be included in the technical docs)
- The other Cycles devs for feedback on the code, especially Sergey for
  mentoring the GSoC project and Brecht for the code review!
- And of course the users who helped with testing, reported bugs and things
  that could and/or should work better!

12 files changed, 19 insertions, 11 deletions

diff --git a/intern/cycles/kernel/closure/bsdf.h b/intern/cycles/kernel/closure/bsdf.h
index a6bba8bf74d..a04c157dc40 100644
--- a/intern/cycles/kernel/closure/bsdf.h
+++ b/intern/cycles/kernel/closure/bsdf.h
@@ -435,5 +435,23 @@ ccl_device bool bsdf_merge(ShaderClosure *a, ShaderClosure *b)
 #endif
 }
 
+/* Classifies a closure as diffuse-like or specular-like.
+ * This is needed for the denoising feature pass generation,
+ * which are written on the first bounce where more than 25%
+ * of the sampling weight belongs to diffuse-line closures. */
+ccl_device_inline bool bsdf_is_specular_like(ShaderClosure *sc)
+{
+	if(CLOSURE_IS_BSDF_TRANSPARENT(sc->type)) {
+		return true;
+	}
+
+	if(CLOSURE_IS_BSDF_MICROFACET(sc->type)) {
+		MicrofacetBsdf *bsdf = (MicrofacetBsdf*) sc;
+		return (bsdf->alpha_x*bsdf->alpha_y <= 0.075f*0.075f);
+	}
+
+	return false;
+}
+
 CCL_NAMESPACE_END
 
diff --git a/intern/cycles/kernel/closure/bsdf_ashikhmin_velvet.h b/intern/cycles/kernel/closure/bsdf_ashikhmin_velvet.h
index 7e0f5a7ec75..a5ba2cb2972 100644
--- a/intern/cycles/kernel/closure/bsdf_ashikhmin_velvet.h
+++ b/intern/cycles/kernel/closure/bsdf_ashikhmin_velvet.h
@@ -40,7 +40,6 @@ typedef ccl_addr_space struct VelvetBsdf {
 
 	float sigma;
 	float invsigma2;
-	float3 N;
 } VelvetBsdf;
 
 ccl_device int bsdf_ashikhmin_velvet_setup(VelvetBsdf *bsdf)
diff --git a/intern/cycles/kernel/closure/bsdf_diffuse.h b/intern/cycles/kernel/closure/bsdf_diffuse.h
index dcd187f9305..ec6f1f20996 100644
--- a/intern/cycles/kernel/closure/bsdf_diffuse.h
+++ b/intern/cycles/kernel/closure/bsdf_diffuse.h
@@ -37,7 +37,6 @@ CCL_NAMESPACE_BEGIN
 
 typedef ccl_addr_space struct DiffuseBsdf {
 	SHADER_CLOSURE_BASE;
-	float3 N;
 } DiffuseBsdf;
 
 /* DIFFUSE */
diff --git a/intern/cycles/kernel/closure/bsdf_diffuse_ramp.h b/intern/cycles/kernel/closure/bsdf_diffuse_ramp.h
index 2d982a95fe4..24f40af46a3 100644
--- a/intern/cycles/kernel/closure/bsdf_diffuse_ramp.h
+++ b/intern/cycles/kernel/closure/bsdf_diffuse_ramp.h
@@ -40,7 +40,6 @@ CCL_NAMESPACE_BEGIN
 typedef ccl_addr_space struct DiffuseRampBsdf {
 	SHADER_CLOSURE_BASE;
 
-	float3 N;
 	float3 *colors;
 } DiffuseRampBsdf;
 
diff --git a/intern/cycles/kernel/closure/bsdf_microfacet.h b/intern/cycles/kernel/closure/bsdf_microfacet.h
index 3fe7572e4ce..30cc8b90330 100644
--- a/intern/cycles/kernel/closure/bsdf_microfacet.h
+++ b/intern/cycles/kernel/closure/bsdf_microfacet.h
@@ -46,7 +46,6 @@ typedef ccl_addr_space struct MicrofacetBsdf {
 	float alpha_x, alpha_y, ior;
 	MicrofacetExtra *extra;
 	float3 T;
-	float3 N;
 } MicrofacetBsdf;
 
 /* Beckmann and GGX microfacet importance sampling. */
diff --git a/intern/cycles/kernel/closure/bsdf_microfacet_multi.h b/intern/cycles/kernel/closure/bsdf_microfacet_multi.h
index 57f1e733ee7..30644946840 100644
--- a/intern/cycles/kernel/closure/bsdf_microfacet_multi.h
+++ b/intern/cycles/kernel/closure/bsdf_microfacet_multi.h
@@ -42,7 +42,7 @@ ccl_device_forceinline float D_ggx_aniso(const float3 wm, const float2 alpha)
 /* Sample slope distribution (based on page 14 of the supplemental implementation). */
 ccl_device_forceinline float2 mf_sampleP22_11(const float cosI, const float2 randU)
 {
-	if(cosI > 0.9999f || cosI < 1e-6f) {
+	if(cosI > 0.9999f || fabsf(cosI) < 1e-6f) {
 		const float r = sqrtf(randU.x / max(1.0f - randU.x, 1e-7f));
 		const float phi = M_2PI_F * randU.y;
 		return make_float2(r*cosf(phi), r*sinf(phi));
diff --git a/intern/cycles/kernel/closure/bsdf_oren_nayar.h b/intern/cycles/kernel/closure/bsdf_oren_nayar.h
index cb342a026ef..6b770fc0c16 100644
--- a/intern/cycles/kernel/closure/bsdf_oren_nayar.h
+++ b/intern/cycles/kernel/closure/bsdf_oren_nayar.h
@@ -22,7 +22,6 @@ CCL_NAMESPACE_BEGIN
 typedef ccl_addr_space struct OrenNayarBsdf {
 	SHADER_CLOSURE_BASE;
 
-	float3 N;
 	float roughness;
 	float a;
 	float b;
diff --git a/intern/cycles/kernel/closure/bsdf_phong_ramp.h b/intern/cycles/kernel/closure/bsdf_phong_ramp.h
index e152a8780db..420f94755ee 100644
--- a/intern/cycles/kernel/closure/bsdf_phong_ramp.h
+++ b/intern/cycles/kernel/closure/bsdf_phong_ramp.h
@@ -40,7 +40,6 @@ CCL_NAMESPACE_BEGIN
 typedef ccl_addr_space struct PhongRampBsdf {
 	SHADER_CLOSURE_BASE;
 
-	float3 N;
 	float exponent;
 	float3 *colors;
 } PhongRampBsdf;
diff --git a/intern/cycles/kernel/closure/bsdf_principled_diffuse.h b/intern/cycles/kernel/closure/bsdf_principled_diffuse.h
index 8a116693bdb..215c32e1ffb 100644
--- a/intern/cycles/kernel/closure/bsdf_principled_diffuse.h
+++ b/intern/cycles/kernel/closure/bsdf_principled_diffuse.h
@@ -28,7 +28,6 @@ typedef ccl_addr_space struct PrincipledDiffuseBsdf {
 	SHADER_CLOSURE_BASE;
 
 	float roughness;
-	float3 N;
 } PrincipledDiffuseBsdf;
 
 ccl_device float3 calculate_principled_diffuse_brdf(const PrincipledDiffuseBsdf *bsdf,
diff --git a/intern/cycles/kernel/closure/bsdf_principled_sheen.h b/intern/cycles/kernel/closure/bsdf_principled_sheen.h
index 58df4f7ddbb..f4476bfecd0 100644
--- a/intern/cycles/kernel/closure/bsdf_principled_sheen.h
+++ b/intern/cycles/kernel/closure/bsdf_principled_sheen.h
@@ -26,7 +26,6 @@ CCL_NAMESPACE_BEGIN
 
 typedef ccl_addr_space struct PrincipledSheenBsdf {
 	SHADER_CLOSURE_BASE;
-	float3 N;
 } PrincipledSheenBsdf;
 
 ccl_device float3 calculate_principled_sheen_brdf(const PrincipledSheenBsdf *bsdf,
diff --git a/intern/cycles/kernel/closure/bsdf_toon.h b/intern/cycles/kernel/closure/bsdf_toon.h
index 28e775bcbc8..d8b6d8ddead 100644
--- a/intern/cycles/kernel/closure/bsdf_toon.h
+++ b/intern/cycles/kernel/closure/bsdf_toon.h
@@ -38,7 +38,6 @@ CCL_NAMESPACE_BEGIN
 typedef ccl_addr_space struct ToonBsdf {
 	SHADER_CLOSURE_BASE;
 
-	float3 N;
 	float size;
 	float smooth;
 } ToonBsdf;
diff --git a/intern/cycles/kernel/closure/bssrdf.h b/intern/cycles/kernel/closure/bssrdf.h
index f9236a6e52c..f733ea4c517 100644
--- a/intern/cycles/kernel/closure/bssrdf.h
+++ b/intern/cycles/kernel/closure/bssrdf.h
@@ -28,7 +28,6 @@ typedef ccl_addr_space struct Bssrdf {
 	float texture_blur;
 	float albedo;
 	float roughness;
-	float3 N;
 } Bssrdf;
 
 /* Planar Truncated Gaussian