**Final project proposal**
(#) dhawals1939 - f006yzt
(##) Features Implemented
1. Parallelization of the Code (2 pts).
2. Intel Open Denoise (3 pts)
3. Stratified Sampling (3 pts).
4. Photon Mapping (8 pts)
(##) Parallelization of the rendering
Observed a `8x` speedup with 8 threads.
(##) Denoising with various parameters
High(4096) SPP vs 100 SPP
High(4096) SPP vs 100 SPP with Intel Open Denoise with Radiance
High(4096) SPP vs 100 SPP with Intel Open Denoise with Color and Normal
High(4096) SPP vs 100 SPP with Intel Open Denoise with Albedo and Normal without clean AUX
Best(with Auxilary Albedo and Normal) vs Simple Radiance Denoise
Denoising with Normal and Albedo assuming both are clean Auxilary data gives the best results. Just using radiance gives
a lot of artifacts image.
(##) Stratified Sampling
Low SPP Comparision
9 SPP Comprarision
36 SPP Comparision
Stratified sampling divides the space into strata and then adds jitter to the samples. This makes the coverage of the
space more uniform. And the renders are of better quality.
I have used 9 and 36 samples per pixel for the comparison. As it needs samples to be square.
I also used the stratification for the lens sample instead of picking one from the random2f instead of just 2D image.
(##) Photon Mapping
Indirect Illumination with light facing upwards.
Once Photon Mapping is done and Photons are visualized
Path tracing vs Photon Mapping
Length of path 4
For the Path Tracer the length of the path is 4 and for the Photon Mapping the length of the path is 2 + 2.
Where 2 is from the photon map generatation shot from the light and 2 is from the forward pass.
Final Scene Comparison with 16spp
- Path Tracer with 200 bounces
- Photon Mapping with 2+2 bounces
Equal Time Comparison \~ 40 seconds
Final Submision
Comparision between blender vs my implementation
Comparison against Path traced with 20 bounces 100 SPP
Modifying the photon map with a beam search
References and help
```
Mitsuba 0.6
Pbrt v3 book
Help from the friends Micheal and Matthew
https://poly.pizza/m/cEbi0acnY23
```