monte carol integration
play

Monte Carol Integration Sung-Eui Yoon ( ) Course URL: - PowerPoint PPT Presentation

CS580: Monte Carol Integration Sung-Eui Yoon ( ) Course URL: http://sglab.kaist.ac.kr/~sungeui/GCG Class Objectives Sampling approach for solving the rendering equation Monte Carlo integration Estimator and its variance


  1. CS580: Monte Carol Integration Sung-Eui Yoon ( 윤성의 ) Course URL: http://sglab.kaist.ac.kr/~sungeui/GCG

  2. Class Objectives ● Sampling approach for solving the rendering equation ● Monte Carlo integration ● Estimator and its variance ● Sampling according to the pdf 2

  3. Two Forms of the Rendering Equation ● Hemisphere integration ● Area integration 3

  4. Radiance Evaluation ● Fundamental problem in GI algorithm ● Evaluate radiance at a given surface point in a given direction ● I nvariance defines radiance everywhere else 4

  5. Radiance Evaluation 5

  6. Why Monte Carlo? ● Radiace is hard to evaluate From kavita’s slides ● Sample many paths ● I ntegrate over all incoming directions ● Analytical integration is difficult ● Need numerical techniques 6

  7. Monte Carlo Integration ● Numerical tool to evaluate integrals ● Use sampling ● Stochastic errors ● Unbiased ● On average, we get the right answer 7

  8. 8

  9. 9

  10. 10

  11. 11

  12. 12

  13. 13

  14. 14

  15. 15

  16. 16

  17. 17

  18. 18

  19. 19

  20. 20

  21. 21

  22. 22

  23. • Consider p(x) for estimate • We will study it as importance sampling later 23

  24. 24

  25. 25

  26. 26

  27. 27

  28. MC Integration - Example ● I ntegral ● Variance 28

  29. 29

  30. 30

  31. 31

  32. Advantages of MC 1 ● Convergence rate of O ( ) N ● Simple ● Sampling ● Point evaluation ● General ● Works for high dimensions ● Deals with discontinuities, crazy functions, etc. 32

  33. 33

  34. 34

  35. Importance Sampling ● Take more samples in important regions, where the function is large From kavita’s slides 35

  36. 36

  37. 37

  38. 38

  39. 39

  40. Sampling according to pdf ● I nverse cumulative distribution function ● Rejection sampling 40

  41. Inverse Cumulative Distribution Function – Discrete Case , given uniform sampling 41

  42. Continuous Random Variable ● Algorithm ● Pick u uniformly from [0, 1)   y  ● Output y = P -1 (u), where P ( y ) p ( x ) dx  42

  43. 43

  44. 44

  45. 45

  46. 46

  47. Rejection Method ● Often not possible to compute the inverse of cdf From kavita’s slides 1 47

  48. Class Objectives were: ● Sampling approach for solving the rendering equation ● Monte Carlo integration ● Estimator and its variance ● Sampling according to the pdf 48

  49. Next Time ● Monte Carlo ray tracing 49

Recommend


More recommend