Sampling Analysis Cengiz ztireli using Correlations Gurprit Singh - - PowerPoint PPT Presentation

sampling analysis
SMART_READER_LITE
LIVE PREVIEW

Sampling Analysis Cengiz ztireli using Correlations Gurprit Singh - - PowerPoint PPT Presentation

Sampling Analysis Cengiz ztireli using Correlations Gurprit Singh for Monte Carlo Rendering Point Patterns in Computer Graphics Random distributions of points with characteristics Fundamental for many applications in graphics Imaging


slide-1
SLIDE 1

Sampling Analysis using Correlations

for Monte Carlo Rendering

Cengiz Öztireli Gurprit Singh

slide-2
SLIDE 2

Point Patterns in Computer Graphics

Random distributions of points with characteristics Fundamental for many applications in graphics

slide-3
SLIDE 3

Imaging

slide-4
SLIDE 4

Patterns of Nature

slide-5
SLIDE 5

Dynamic Structures

slide-6
SLIDE 6

Simulations

slide-7
SLIDE 7

Geometry Processing

slide-8
SLIDE 8

Fabrication

slide-9
SLIDE 9

Non-photorealistic Rendering

slide-10
SLIDE 10

Rendering – Computing Integrals

slide-11
SLIDE 11

Estimating Integrals with Points

Sample and sum the sampled values of an integrand I := 1 |D| Z

D

f(x)dx ˆ I :=

n

X

I=1

wif(xi) biasP[ˆ I] = I − EP[ˆ I] varP[ˆ I] = EP[ˆ I2] − (EP[ˆ I])2

slide-12
SLIDE 12

Stochastic Point Processes

Formal characterization of point patterns

slide-13
SLIDE 13

Stochastic Point Processes

Formal characterization of point patterns

Point Process

slide-14
SLIDE 14

Stochastic Point Processes

Examples of point processes

Natural Process Manuel Process

slide-15
SLIDE 15

General Point Processes

Infinite point processes

Observation window

slide-16
SLIDE 16

General Point Processes

Assign a random variable to each set

B

NpBq “ 3

B

NpBq “ 5

B

NpBq “ 2

slide-17
SLIDE 17

General Point Processes

Joint probabilities define the point process

B1 B1 B1 B2 B2 B2

pNpB1q,NpB2q

slide-18
SLIDE 18

Point Process Statistics

Correlations as probabilities

Product density Small volumes Points in space

xi dVi

%(n)(x1, · · · , xn)dV1 · · · dVn = p(x1, · · · , xnq

slide-19
SLIDE 19

Point Process Statistics

First order product density

x

Expected number of points around x Measures local density

%(1)(x) = (x)

slide-20
SLIDE 20

Point Process Statistics

First order product density

) = (x)

slide-21
SLIDE 21

Point Process Statistics

First order product density

) = (x)

Constant

slide-22
SLIDE 22

Expected number of points around x & y Measures the joint probability p(x, y)

x y %(2)(x, y) = %(x, y)

Point Process Statistics

Second order product density

slide-23
SLIDE 23

Expected number of points around x, y, z

x y

Point Process Statistics

Higher order product density?

z

Not necessary: second order dogma

slide-24
SLIDE 24

Point Process Statistics

Higher order not necessary: second order dogma

%(2)(x, y) = %(x, y) %(1)(x) = (x) x x y

slide-25
SLIDE 25

Point Process Statistics

Summary: 1st & 2nd order correlations sufficient

%(2)(x, y) = %(x, y) %(1)(x) = (x) x x y

slide-26
SLIDE 26

Point Process Statistics

Example: homogenous Poisson point process a.k.a. random sampling

p(x) = p p(x, y) = p(x)p(y) p(x, y) = %(x, y)dVxdVy λ(x)dV = p λ(x) = λ

y = p(x)p(y) =

) = (x)dVx(y)dVy %(x, y) = (x)(y) = 2

slide-27
SLIDE 27

Point Process Statistics

Summary: 1st & 2nd order correlations sufficient

%(2)(x, y) = %(x, y) %(1)(x) = (x) x x y

slide-28
SLIDE 28

Stationary Point Processes

Stationary (translation invariant) Isotropic (translation & rotation invariant)

slide-29
SLIDE 29

Stationary Point Processes

Stationary (translation invariant)

λ(x) = λ %(x, y) = %(x − y) = λ2g(x − y)

Pair Correlation Function (PCF) DoF reduced from 2d to d

slide-30
SLIDE 30

Stationary Point Processes

Isotropic point process (translation & rotation invariant)

λ(x) = λ g(x − y) = g(||x − y||)

PCF

r

gprq

slide-31
SLIDE 31

EP 2 4X

I6=j

f(xi, xj) 3 5 = Z

Rd⇥Rd f(x, y)%(x, y)dxdy

(xi

Estimating Correlations

Campbell’s Theorem

EP hX f(xi) i = Z

Rd f(x)λ(x)dx

slide-32
SLIDE 32

Estimating Correlations

First order EP hX ID(xi) i = EP " X

xi∈D

1 # # = Z

D

λdx = λ Z

D

dx = λ|D| ˆ λ = P

Pk Nk(D)

K|D|

λ(x) =

Point distribution Number of point distributions

slide-33
SLIDE 33

Estimating Correlations

Second order stationary - pair correlation function (PCF)

EP 2 4X

I6=j

δ(r − (xi − xj)) 3 5 = Z

Rd×Rd (r − (x − y))%(x − y)dxdy

= λ2 Z

Rd×Rd δ(r − (x − y))g(x − y)dxdy = λ2g(r)

(xi

slide-34
SLIDE 34

ˆ g(r) = 1 Kλ2aID(r) X

Pk

X

xi,xj2Pk,i6=j

δ(r − (xi − xj))

Estimating Correlations

Second order stationary - pair correlation function (PCF) Finite domains: ˆ g(r) = 1 Kλ2 X

Pk

X

xi,xj2Pk,i6=j

δ(r − (xi − xj))

slide-35
SLIDE 35

Estimating Correlations

Second order stationary - pair correlation function (PCF)

Pair Correlation Function Point Distribution

slide-36
SLIDE 36

ˆ g(r) = 1 λ2rd1|Sd| X

i6=j

k(r kxi xjk) Estimating Correlations

Second order isotropic - pair correlation function (PCF)

Kernel e.g. Gaussian Volume of the unit hypercube in d dimensions

slide-37
SLIDE 37

Pair Correlation Function

1

ˆ g(r) = ˆ g(r) =

slide-38
SLIDE 38

Pair Correlation Function ˆ g(r) =

1

ˆ g(r) = ˆ g(r) =

slide-39
SLIDE 39

Pair Correlation Function

1

ˆ g(r) = ˆ g(r) =

slide-40
SLIDE 40

Spectral Statistics

Power spectrum Fourier transform

  • f PCF
slide-41
SLIDE 41

Spectral Statistics

PCF Power spectrum Points Points

slide-42
SLIDE 42

Spectral Statistics

Power spectrum Radial average Radial anisotropy

slide-43
SLIDE 43

Statistics for Stationary Processes

Summary Stationary: Spatial (PCF) & spectral (power spectrum) Isotropic: radial averages

PCF Power spectrum