Comparison of light sampling techniques and their combinations in the Dining room scene
Comparison of different combinations of standard light sampling techniques (Light, Spherical) and a new one (Parallel) motivated by properties of the optimal MIS weights. The combination with the new technique using the optimal weights performs best. See Section 8.3 in the main paper for discussion.
Compared variants
| Variant | Used shortcut |
|---|
| Reference | Ref |
| Spherical | S |
| UniformArea | U |
| Parallel | P |
| Spherical + UniformArea, balance heuristic | S+U bal |
| Spherical + UniformArea, power heuristic | S+U pow |
| Spherical + UniformArea, optimal weights - direct estimator | S+U opt-dir |
| Spherical + UniformArea, optimal weights - progressive etimator with update step U | S+U opt-prog-U |
| Spherical + Parallel, balance heuristic | S+P bal |
| Spherical + Parallel, power heuristic | S+P pow |
| Spherical + Parallel, optimal weights - direct estimator | S+P opt-dir |
| Spherical + Parallel, optimal weights - progressive etimator with update step U | S+P opt-prog-U |
Equal-time comparison
Number of iterations is set so each variant takes approximately 20 s to render.
| Variant |
Time [s] |
Iterations |
MSE |
Speedup (to S) |
Speedup (to S+P pow) |
| S |
19.76 |
62 |
1.81E-004 |
1.00 |
2.93 |
| U |
20.65 |
68 |
1.10E-003 |
0.16 |
0.48 |
| P |
19.64 |
63 |
1.64E-003 |
0.11 |
0.32 |
| S+U bal |
20.30 |
47 |
3.12E-004 |
0.58 |
1.70 |
| S+P bal |
19.58 |
47 |
4.69E-004 |
0.39 |
1.13 |
| S+U pow |
19.94 |
47 |
4.18E-004 |
0.43 |
1.27 |
| S+P pow |
20.13 |
48 |
5.31E-004 |
0.34 |
1.00 |
| S+U opt-dir |
19.14 |
44 |
9.49E-005 |
1.91 |
5.59 |
| S+P opt-dir |
19.54 |
45 |
5.31E-005 |
3.40 |
9.99 |
| S+U opt-prog-1 |
20.28 |
35 |
1.85E-004 |
0.98 |
2.86 |
| S+U opt-prog-2 |
20.40 |
40 |
1.26E-004 |
1.43 |
4.21 |
| S+U opt-prog-3 |
19.80 |
40 |
1.31E-004 |
1.38 |
4.04 |
| S+U opt-prog-4 |
19.82 |
41 |
1.35E-004 |
1.34 |
3.94 |
| S+U opt-prog-5 |
20.05 |
42 |
1.38E-004 |
1.31 |
3.86 |
| S+U opt-prog-6 |
19.89 |
42 |
1.44E-004 |
1.26 |
0.00 |
| S+U opt-prog-7 |
19.87 |
42 |
1.48E-004 |
1.22 |
3.58 |
| S+U opt-prog-8 |
20.14 |
43 |
1.51E-004 |
1.20 |
3.52 |
| S+U opt-prog-9 |
19.53 |
42 |
1.63E-004 |
1.11 |
3.26 |
| S+U opt-prog-10 |
20.06 |
43 |
1.64E-004 |
1.10 |
3.24 |
| S+P opt-prog-1 |
20.41 |
36 |
9.69E-005 |
1.87 |
5.48 |
| S+P opt-prog-2 |
20.26 |
40 |
9.68E-005 |
1.87 |
5.48 |
| S+P opt-prog-3 |
19.49 |
40 |
1.08E-004 |
1.68 |
4.92 |
| S+P opt-prog-4 |
19.51 |
41 |
1.21E-004 |
1.50 |
4.40 |
| S+P opt-prog-5 |
20.29 |
43 |
1.24E-004 |
1.46 |
4.27 |
| S+P opt-prog-6 |
20.10 |
43 |
1.36E-004 |
1.33 |
3.91 |
| S+P opt-prog-7 |
19.41 |
42 |
1.53E-004 |
1.18 |
3.47 |
| S+P opt-prog-8 |
19.15 |
41 |
1.72E-004 |
1.05 |
3.09 |
| S+P opt-prog-9 |
19.89 |
43 |
1.70E-004 |
1.06 |
3.12 |
| S+P opt-prog-10 |
20.21 |
44 |
1.74E-004 |
1.04 |
3.04 |
Equal-sample comparison
Each variant renders 20 iterations, i.e. draws 20 samples per technique per pixel. For the single-technique variants results with the same total number of samples per pixel (i.e. 40) are also included.
| Variant |
Time [s] |
Iterations |
MSE |
Overhead (to S 40) |
Overhead (to S+P pow) |
Improvement (to S) |
Improvement (to S+P pow) |
| S |
6,54 |
20 |
5,66E-004 |
-48,89% |
-22,07% |
0,48 |
2,29 |
| S 40 |
12,79 |
40 |
2,73E-004 |
0,00% |
52,49% |
1,00 |
4,74 |
| U |
5,98 |
20 |
3,79E-003 |
-53,27% |
-28,75% |
0,07 |
0,34 |
| U 40 |
12,13 |
40 |
1,88E-003 |
-5,14% |
44,65% |
0,15 |
0,69 |
| P |
6,22 |
20 |
5,22E-003 |
-51,34% |
-25,80% |
0,05 |
0,25 |
| P 40 |
12,46 |
40 |
2,60E-003 |
-2,61% |
48,50% |
0,10 |
0,50 |
| S+U bal |
8,53 |
20 |
7,45E-004 |
-33,28% |
1,74% |
0,37 |
1,74 |
| S+P bal |
8,35 |
20 |
1,13E-003 |
-34,73% |
-0,47% |
0,24 |
1,15 |
| S+U pow |
8,55 |
20 |
9,96E-004 |
-33,15% |
1,94% |
0,27 |
1,30 |
| S+P pow |
8,39 |
20 |
1,29E-003 |
-34,42% |
0,00% |
0,21 |
1,00 |
| S+U opt-dir |
8,82 |
20 |
2,19E-004 |
-31,00% |
5,21% |
1,25 |
5,91 |
| S+P opt-dir |
8,89 |
20 |
1,29E-004 |
-30,53% |
5,94% |
2,12 |
10,05 |
| S+U opt-prog-1 |
11,34 |
20 |
3,84E-004 |
-11,31% |
35,25% |
0,71 |
3,37 |
| S+U opt-prog-2 |
10,18 |
20 |
2,99E-004 |
-20,43% |
21,33% |
0,91 |
4,32 |
| S+U opt-prog-3 |
9,84 |
20 |
3,20E-004 |
-23,07% |
17,31% |
0,85 |
4,04 |
| S+U opt-prog-4 |
9,62 |
20 |
3,48E-004 |
-24,75% |
14,75% |
0,79 |
3,72 |
| S+U opt-prog-5 |
9,46 |
20 |
3,72E-004 |
-26,04% |
12,78% |
0,73 |
3,48 |
| S+U opt-prog-6 |
9,54 |
20 |
4,00E-004 |
-25,40% |
13,76% |
0,68 |
3,23 |
| S+U opt-prog-7 |
9,33 |
20 |
4,25E-004 |
-27,09% |
11,18% |
0,64 |
3,04 |
| S+U opt-prog-8 |
9,46 |
20 |
4,53E-004 |
-26,02% |
12,81% |
0,60 |
2,85 |
| S+U opt-prog-9 |
9,63 |
20 |
4,74E-004 |
-24,73% |
14,77% |
0,58 |
2,73 |
| S+U opt-prog-10 |
9,25 |
20 |
5,04E-004 |
-27,66% |
10,31% |
0,54 |
2,56 |
| S+P opt-prog-1 |
11,23 |
20 |
2,15E-004 |
-12,17% |
33,92% |
1,27 |
6,00 |
| S+P opt-prog-2 |
10,18 |
20 |
2,65E-004 |
-20,43% |
21,33% |
1,03 |
4,88 |
| S+P opt-prog-3 |
9,72 |
20 |
3,15E-004 |
-23,98% |
15,92% |
0,87 |
4,10 |
| S+P opt-prog-4 |
9,45 |
20 |
3,69E-004 |
-26,09% |
12,70% |
0,74 |
3,50 |
| S+P opt-prog-5 |
9,46 |
20 |
4,20E-004 |
-26,06% |
12,75% |
0,65 |
3,08 |
| S+P opt-prog-6 |
9,32 |
20 |
4,72E-004 |
-27,11% |
11,15% |
0,58 |
2,74 |
| S+P opt-prog-7 |
9,32 |
20 |
5,21E-004 |
-27,16% |
11,08% |
0,52 |
2,48 |
| S+P opt-prog-8 |
9,31 |
20 |
5,67E-004 |
-27,17% |
11,05% |
0,48 |
2,28 |
| S+P opt-prog-9 |
9,32 |
20 |
6,12E-004 |
-27,16% |
11,06% |
0,45 |
2,11 |
| S+P opt-prog-10 |
9,06 |
20 |
6,65E-004 |
-29,13% |
8,06% |
0,41 |
1,94 |
