Comparison of light selection techniques and their combinations in the Staircase I scene
Comparison of different combination strategies for a trained light selection technique (Trained) and a defensive technique (Uniform). In contrast to the balance/power heuristic, the optimal MIS weights are never worse than any of the two techniques alone. See Section 8.3 in the main paper for discussion.
Compared variants
| Variant | Used shortcut |
|---|
| Reference | Ref |
| Uniform | U |
| Trained | T |
| Trained + Uniform, balance heuristic | T+U bal |
| Trained + Uniform, power heuristic | T+U pow |
| Trained + Uniform, optimal weights - direct etimator | T+U opt-dir |
| Trained + Uniform, optimal weights - progressive etimator with update step U | T+U 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 T+U pow) |
| U |
19.59 |
29 |
5.60E-009 |
0.27 |
| T |
19.70 |
28 |
1.87E-009 |
0.81 |
| T+U bal |
20.12 |
23 |
2.09E-009 |
0.73 |
| T+U pow |
20.15 |
23 |
1.52E-009 |
1.00 |
| T+U opt-dir |
19.54 |
21 |
1.71E-010 |
8.89 |
| T+U opt-prog-1 |
19.78 |
17 |
5.06E-010 |
3.01 |
| T+U opt-prog-2 |
20.08 |
19 |
5.52E-010 |
2.76 |
| T+U opt-prog-3 |
20.22 |
20 |
6.46E-010 |
2.36 |
| T+U opt-prog-4 |
19.83 |
20 |
7.48E-010 |
2.03 |
| T+U opt-prog-5 |
19.55 |
20 |
8.67E-010 |
1.75 |
| T+U opt-prog-6 |
19.64 |
20 |
9.73E-010 |
1.56 |
| T+U opt-prog-7 |
19.34 |
20 |
1.08E-009 |
1.40 |
| T+U opt-prog-8 |
19.33 |
20 |
1.18E-009 |
1.29 |
| T+U opt-prog-9 |
19.28 |
20 |
1.29E-009 |
1.18 |
| T+U opt-prog-10 |
20.29 |
21 |
1.27E-009 |
1.19 |
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 T+U pow) |
Improvement (to T+U pow) |
| U |
13.57 |
20 |
8.08E-009 |
-23.02% |
0.22 |
| U 40 |
26.96 |
40 |
4.01E-009 |
52.97% |
0.43 |
| T |
14.21 |
20 |
2.62E-009 |
-19.39% |
0.66 |
| T 40 |
28.03 |
40 |
1.32E-009 |
59.02% |
1.32 |
| T+U bal |
17.56 |
20 |
2.39E-009 |
-0.40% |
0.73 |
| T+U pow |
17.63 |
20 |
1.74E-009 |
0.00% |
1.00 |
| T+U opt-dir |
18.72 |
20 |
1.82E-010 |
6.20% |
9.56 |
| T+U opt-prog-1 |
23.45 |
20 |
3.99E-010 |
33.02% |
4.37 |
| T+U opt-prog-2 |
21.03 |
20 |
5.09E-010 |
19.32% |
3.42 |
| T+U opt-prog-3 |
20.29 |
20 |
6.46E-010 |
15.14% |
2.70 |
| T+U opt-prog-4 |
19.82 |
20 |
7.48E-010 |
12.44% |
2.33 |
| T+U opt-prog-5 |
19.67 |
20 |
8.67E-010 |
11.57% |
2.01 |
| T+U opt-prog-6 |
19.66 |
20 |
9.73E-010 |
11.53% |
1.79 |
| T+U opt-prog-7 |
19.37 |
20 |
1.08E-009 |
9.92% |
1.61 |
| T+U opt-prog-8 |
19.34 |
20 |
1.18E-009 |
9.70% |
1.48 |
| T+U opt-prog-9 |
19.45 |
20 |
1.29E-009 |
10.35% |
1.35 |
| T+U opt-prog-10 |
19.14 |
20 |
1.38E-009 |
8.56% |
1.26 |
