GR2 Advanced Computer Graphics AGR - PowerPoint PPT Presentation

1 / 25
About This Presentation
Title:

GR2 Advanced Computer Graphics AGR

Description:

... chosen, then repeatedly refined until deemed to converge ... and we can continue until the iteration converges. This is known as the Jacobi iterative method ... – PowerPoint PPT presentation

Number of Views:16
Avg rating:3.0/5.0
Slides: 26
Provided by: kenbr
Category:

less

Transcript and Presenter's Notes

Title: GR2 Advanced Computer Graphics AGR


1
GR2Advanced Computer GraphicsAGR
  • Lecture 15
  • Radiosity

2
Review
  • First a review of the two rendering approaches we
    have studied
  • Phong reflection model
  • ray tracing

3
Phong Reflection Model
  • This is the most common approach to rendering
  • objects represented as polygonal faces
  • intensity of faces calculated by Phong local
    illumination model
  • polygons projected to viewplane
  • Gouraud shading applied with Z buffer to
    determine visibility

4
Phong Reflection Model
  • Strengths
  • simple and efficient
  • models ambient, diffuse and specular reflection
  • Limitations
  • only considers light incident from a light
    source, and not inter-object reflections - ie it
    is a local illumination method (ambient term is
    approximation to global illumination
  • empirical rather than theoretical base
  • objects typically have plastic appearance

5
Ray Tracing
  • Ray traced from viewpoint through pixel until
    first object intersected
  • Colour calculated as summation of
  • local Phong reflection at that point
  • specularly reflected light from direction of
    reflection
  • transmitted light from refraction direction if
    transparent

6
Ray Tracing
  • This is done recursively
  • Colour of light incoming along reflection
    direction found by
  • tracing ray back until it hits an object
  • colour of light emitted by object is itself
    summation of local component, reflected component
    and transmitted component
  • and so on

7
Ray Tracing - Strengths and Weaknesses
  • Advantages
  • increased realism through ability to handle
    inter-object reflection
  • Disadvantages
  • much more expensive than local reflection
  • still empirical
  • only handles specular inter-object reflection
  • entire calculation is view-dependent

8
Radiosity
  • Based on the physics of heat transfer between
    surfaces
  • Developed in 1980s at Cornell University in US
    (Cohen, Greenberg)
  • Determine energy balance of light transfer
    between all surfaces in an enclosed space
  • equilibrium reached between emission of light and
    partial absorption of light
  • Assume surfaces are opaque, are perfect diffuse
    reflectors and are represented as sets of
    rectangular patches Ai

9
Radiosity Examples
10
Radiosity - Definition
  • Radiosity defined as
  • energy (Bi ) per unit area leaving a surface
    patch (Ai ) per unit time

Bi Ai Ei Ai Ri ? ( Fj-i Bj Aj) i1,2,..N
j
light emitted
light reflected
light leaving
Ei is the light energy emitted by Ai per unit
area Ri is the fraction of incident light
reflected in all directions Fj-i is the fraction
of energy leaving Aj that reaches Ai
11
Radiosity - Pictorial Definition
Form factor Fj-i is fraction of energy leaving Aj
that reaches Ai. It is determined by
the relative orientation of the patches and
distance r between them
Aj
?j
r
Nj
Ni
Form factors are hard to calculate! Lets assume
for now we can do it.
?i
Ai
Bi Ai Ei Ai Ri ? (Fj-i Bj Aj)
j
12
Simplifying the Equation
Bi Ai Ei Ai Ri ? (Fj-i Bj Aj) ie Bi Ai Ei
Ai Ri ? (Bj Fj-i Aj)
j
j
There is a reciprocity relationship Fj-i Aj
Fi-jAi
Hence
Bi Ai Ei Ai Ri ? (Bj Fi-j Ai)
j
So the radiosity of patch Ai is given by
Bi Ei Ri ? ( Bj Fi-j )
j
13
Creating a System of Equations
  • We get one equation for each patch
  • assume we can calculate form factors
  • then N equations for N unknowns B1,B2,..BN
  • First equation is
  • (1-R1F1-1)B1 - (R1F1-2)B2 - (R1F1-3)B3 - .. -
    (R1F1-N)BN E1
  • ..and we get in all N equations like this.
  • Generally Fi-i will be zero - why?
  • Most of the Ei will be zero - why?

14
Solving the Equations
  • Equations are solved iteratively - ie a first
    guess chosen, then repeatedly refined until
    deemed to converge
  • Suppose we guess solution as
  • B1(0), B2(0), .. BN(0)
  • Rewrite first equation as
  • B1 E1 (R1F1-2)B2 .. (R1F1-N)BN
  • Then we can improve estimate of B1 by
  • B1(1) E1 (R1F1-2)B2(0) .. (R1F1-N)BN(0)
  • .. and so on for other Bi

15
Solving the Equations
  • This gives an improved estimate
  • B1(1), B2(1), .. BN(1)
  • and we can continue until the iteration
    converges.
  • This is known as the Jacobi iterative method
  • An improved method is Gauss-Seidel iteration
  • this always uses best available values
  • eg B1(1) (rather than B1(0) ) used to calculate
    B2(1), etc

16
Rendering
  • What have we calculated?
  • The Bi are the intensities of light emanating
    from each patch
  • the form factors do not depend on wavelength ?,
    but the Ri do
  • thus Bi depend on ?, so we need to calculate
    BiRED, BiGREEN, BiBLUE
  • We get vertex intensities by averaging the
    intensities of surrounding faces
  • Then we can pass to Gouraud renderer code for
    interpolated shading

17
Pause at this stage
  • Number of equations can be very large
  • Calculation is not view dependent
  • Only diffuse reflection
  • We still have not seen how to calculate the form
    factors!
  • Their calculation dominates

18
Form Factors
  • The calculation of the form factors is
    unfortunately quite hard
  • We begin by looking at the form factor between
    two infinitesimal areas on the patches

19
Form Factors - Notation
Aj
Form factor Fdi-dj gives the fraction of energy
reaching dAj from dAi.
dAj
Nj
?j
Ni
r
?i
Ai
r is distance between elements Ni, Nj are the
normals ?i, ?j are angles made with normals by
line joining elements
dAi
20
Form Factor Calculation 2D Cross Section View
Draw in 2D but imagine this in 3D create unit
hemisphere with dAi at centre light
emits/reflects equally in all directions from dAi
form factor Fdi-dj is fraction of
energy reaching dAj from dAi
21
Form Factor Calculation 2D Cross Section View
We begin calculation by projecting dAj onto
the surface of the hemisphere (blue) this
resulting area is (cos ?j / r2 ) dAj
Nj
?j
r
This gives us the relative area of light energy
reaching dAj from dAi
22
Form Factor Calculation 2D Cross Section View
But we need a measure per unit area of Ai so need
to adjust for orientation of Ai This gives a
corrected area as ( cos ?i cos ?j / r2 ) dAj
?i
23
Form Factor Calculation 2D Cross Section View
Total energy comes from integrating over whole
hemisphere - this comes to ? Hence form factor
is given by Fdi-dj ( cos ?i cos ?j ) / ( ?
r2 )dAj
dAi
24
Form Factor Calculation 2D Cross Section View
Finally we need to sum up for ALL dAj. This
means integrating over the whole of the
patch Fdi-j ? (cos ?i cos ?j ) /( ? r2 ) dAj
25
Form Factor Calculation
Strictly speaking, we should now integrate over
all dAi. In practice, we take Fdi-j as
representative of Fi-j and this assumption Fi-j
Fdi-j works OK in practice.
However the calculation of the integral ? (cos ?i
cos ?j ) /( ? r2 ) dAj is extremely difficult.
Next lecture will discuss an approximation.
Write a Comment
User Comments (0)
About PowerShow.com