Ray Tracing Basics

CSE 681 Autumn 11

Han-Wei Shen
 Forward Ray Tracing
• We shoot a large number of photons

Problem?
 Backward Tracing
For every pixel
Construct a ray from the eye
For every object in the scene
Find intersection with the ray
Keep if closest
 The Viewing Model
• Based on a simple Pinhole Camera model
 Simplest lens model  Perfect image if hole
 Inverted image infinitely small
 Similar triangles  Pure geometric optics
 No blurry

pin-hole camera simplified pin-hole camera

 Simplified Pinhole Camera
 Eye = pinhole, Image plane = box face (re-arrange)
 Eye-image pyramid (frustum)
 Note that the distance/size of image are arbitrary
 Basic Ray Tracing Algorithm
for every pixel {
cast a ray from the eye
for every object in the scene
find intersections with the ray
keep it if closest
}
compute color at the intersection point
}
 Construct a Ray
3D parametric line
eye
p(t) = eye + t (s-eye) r(t)
p
t=0
r(t): ray equation
eye: eye (camera) position
s: pixel position
t: ray parameter

Question: How to calculate the pixel position P?

 Constructing a Ray
• 3D parametric line
p(t) = e + t (s-e) s
s-e e
*(boldface means vector)

• So we need to know e and s

• What are given (specified by the user or scene

file)?
✓ camera position
✓ camera direction or center of interest
✓ camera orientation or
view up vector
✓ distance to image plane
✓ field of view + aspect ratio
✓ pixel resolution
 Given Camera Information
• Camera
• Eye
• Look at
• Orientation (up vector)
v
• Image plane n
yres
• Distance to plane, N e
• Field of view in Y u
• Aspect ration (X/Y)

xres
• Screen
• Pixel resolution N
Construct Eye Coordinate System
• We can calculate the pixel positions much
more easily if we construct an eye
coordinate system (eye space) first
 Known: eye position, center of interest, view-up
vector
 To find out: new origin and three basis vectors

center of interest (COI)

Assumption: the direction of view is
eye orthogonal to the view plane (the plane
that objects will be projected onto)
 Eye Coordinate System
 Origin: eye position
 Three basis vectors: one is the normal vector (n) of
the viewing plane, the other two are the ones (u and
v) that span the viewing plane

v
u
Center of interest (COI)
eye
n

(u,v,n should be orthogonal to each other)

 Eye Coordinate System
 Origin: eye position
 Three basis vectors: one is the normal vector (n) of
the viewing plane, the other two are the ones (u and
v) that span the viewing plane
n is pointing away from the
v
u world because we use right
hand coordinate system
Center of interest (COI)
eye N = eye – COI
n n= N / |N|

Remember u,v,n should

be all unit vectors

(u,v,n should be orthogonal to each other)

 Eye Coordinate System
 What about u and v?

We can get u first -

v
V_up u
u is a vector that is perpendicular
to the plane spanned by
COI N and view up vector (V_up)
eye
n
 Eye Coordinate System
 What about u and v?

We can get u first -

v
V_up u
u is a vector that is perpendicular
to the plane spanned by
COI N and view up vector (V_up)
eye
n

U = V_up x n

u = U / |U|
 Eye Coordinate System
 What about v?
Knowing n and u, getting v is
easy
v
V_up u

COI
eye
n
 Eye Coordinate System
 What about v?
Knowing n and u, getting v is
easy
v
V_up u

COI
eye v = n xu
n

v is already normalized
 Eye Coordinate System
 Put it all together

v Eye space origin: (Eye.x , Eye.y, Eye.z)

V_up u
Basis vectors:

COI n = (eye – COI) / | eye – COI|

eye
n u = (V_up x n) / | V_up x n |
v = n x u
 Next Step?
• Determine the size of the image plane
• This can be derived from
✓ distance from the camera to the center of
the image plane
✓ Vertical field of view angle
✓ Aspect ratio of the image plane
★ Aspect ratio being Width/Height
 Image Plane Setup
W
• Tan(θv /2) = H / 2d
e
C
• W = H * aspect_ratio H θv

L
• C’s position = e - n * d

• L’s position = C - u * W/2 - v * H/2 d

• Assuming the image resolution is X (horizontal) by Y

(vertical), then each pixel has a width of W/X and a height
of H/Y

• Then for a pixel s at the image pixel (i,j) , it’s location is at

L + u * i * W/X + v * j * H/Y
 Put it all together
• We can represent the ray as a 3D parametric line
p(t) = e + t (s-e)
(now you know how to get s and e)
s
s-e e
• Typically we offset the ray by half
of the pixel width and height, i.e, cast the ray from the pixel
center

incrementing
(i,j)
(0,0)
 Put it all together
• We can represent the ray as a 3D parametric line
p(t) = e + t (s-e)
(now you know how to get s and e)
s
s-e e
• Typically we offset the ray by half
of the pixel width and height, i.e, cast the ray from the pixel
center

incrementing
(i,j)
(0,0)
Ray-Sphere Intersection
• Problem: Intersect a line with a sphere
✓ A sphere with center c = (xc,yc,zc) and radius R can be
represented as:
2 2 2 2
(x-x + (y-yc) + (z-zc) - R = 0
c)

✓ For a point p on the sphere, we can write the above in

vector form:
2
(p-c).(p-c) - R = 0 (note ‘.’ is a dot product)
✓ We can plug the point on the ray p(t) = e + t d
2
(e+td-c).(e+td-c) - R = 0 and yield
2 2
(d.d) t + 2d.(e-c)t + (e-c).(e-c) - R = 0
Ray-Sphere Intersection
• When solving a quadratic equation
2
at + bt + c = 0
We have
• Discriminant
• and Solution
 Ray-Sphere Intersection
b2 – 4ac < 0 ⇒ No intersection
b2 – 4ac > 0 ⇒ Two solutions (enter and exit)
b2 – 4ac = 0 ⇒ One solution (ray grazes sphere)

 Should we use the larger or smaller t value?

 Ray-Sphere Intersection
b2 – 4ac < 0 ⇒ No intersection
b2 – 4ac > 0 ⇒ Two solutions (enter and exit)
b2 – 4ac = 0 ⇒ One solution (ray grazes sphere)

 Should we use the larger or smaller t value?

 Calculate Normal
• Needed for computing lighting
Q = P(t) – C … and remember Q/||Q||

C
t
Q
 Calculate Normal
• Needed for computing lighting
Q = P(t) – C … and remember Q/||Q||

C
t
Q
normal
 Choose the closet sphere
• Minimum search problem
For each pixel {
form ray from eye through the pixel center
tmin = ∞
For each object {
if (t = intersect(ray, object)) {
if (t < tmin) {
closestObject = object
tmin = t
}
}
}
}
 Final Pixel Color
if (tmin == ∞)
pixelColor = background color
else
pixelColor = color of object at d along ray

eye ray object

 CSE 681
Ray-Object Intersections:
Axis-aligned Box
Ray-Box Intersection Test

Y = y2

Z = z2

X = x1 X = x2
Z = z1

Y = y1
Ray-Box Intersection Test

Y = y2

Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
• Intersect ray with each plane
– Box is the union of 6 planes
x = x1, x = x2 Y = y2

y = y1, y = y2
z = z1, z = z2 Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
• Intersect ray with each plane
– Box is the union of 6 planes
x = x1, x = x2 Y = y2

y = y1, y = y2
z = z1, z = z2 Z = z2

X = x1 X = x2
Z = z1
• Ray/axis-aligned plane
is easy: Y = y1
 Ray-Box Intersection Test
• Intersect ray with each plane
– Box is the union of 6 planes
x = x1, x = x2 Y = y2

y = y1, y = y2
z = z1, z = z2 Z = z2

X = x1 X = x2
Z = z1
• Ray/axis-aligned plane
is easy: Y = y1
 Ray-Box Intersection Test
• Intersect ray with each plane
– Box is the union of 6 planes
x = x1, x = x2 Y = y2

y = y1, y = y2
z = z1, z = z2 Z = z2

X = x1 X = x2
Z = z1
• Ray/axis-aligned plane
is easy: Y = y1

E.g., solve x component: ex + tDx = x1

Ray-Box Intersection Test

Y = y2

Z = z2

X = x1 X = x2
Z = z1

Y = y1
Ray-Box Intersection Test

Y = y2

Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
Y = y2

Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

with the smallest t > 0

Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

with the smallest t > 0

that is within the range Z = z2

X = x1 X = x2
Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

with the smallest t > 0

that is within the range Z = z2

X = x1 X = x2
of the box Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

with the smallest t > 0

that is within the range Z = z2

X = x1 X = x2
of the box Z = z1

Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

with the smallest t > 0

that is within the range Z = z2

X = x1 X = x2
of the box Z = z1

• We can do more Y = y1
 Ray-Box Intersection Test
1. Intersect the ray with each plane
2. Sort the intersections
3. Choose intersection Y = y2

with the smallest t > 0

that is within the range Z = z2

X = x1 X = x2
of the box Z = z1

• We can do more Y = y1

efficiently
 Only Consider 2D for Now
• if a point (x,y) is in the box, then (x,y) in
[x1 , x2] x [y1, y2]

y = y1

y = y1

x = x1 x = x2
 The Principle
• Assuming the ray hits the box boundary lines at
intervals [txmin,txmax], [tymin,tymax], the ray
hits the box if and only if the intersection of the
two intervals is not empty
txmax

tymax txmin txmax

txmin
tymin tymax

tymin
28
 Pseudo Code
txmin =(x1 - ex )/Dx
//assume Dx >0
txmax =(x2 - ex )/Dx
tymin = (y1 - ey )/Dy
tymax = (y2 - ey )/Dy //assume Dy >0
if (txmin > tymax) or (tymin > txmax)
return false
else
return true

29
 Pseudo Code

txmin =(x2 - ex )/Dx //if Dx < 0

txmax =(x1 - ex )/Dx
tymin = (y2 - ey )/Dy
//if Dy < 0
tymax = (y1 - ey )/Dy
if (txmin > tymax) or (tymin > txmax)
return false
else
return true

30
6
 Now Consider All Axis
• We will calculate t1 and t2 for each axis (x,
y, and z)
• Update the intersection interval as we
compute t1 and t2 for each axis t2
• remember: t1
t1=(x1- px)/Dx
t2=(x2- px)/Dx
D x = x1 x = x2

p
 Update [tnear, tfar]
• Set tnear = -∞ and tfar = +∞
• For each axis, compute t1 and t2
– make sure t1 < t2
– if t1 > tnear, tnear =t1
t2
– if t2 < tfar, tfar = t2
t1

• If tnear > tfar, box is missed

D x = x1 x = x2

p
 Algorithm
Set tnear = - ∞, tfar = ∞
R(t) = p + t * D
For each pair of planes P associated with X, Y, and Z do: (example uses X
planes)
if direction Dx = 0 then
if (px < x1 or px > x2)
return FALSE
else
begin
t1 = (xl - px) / Dx
t2 = (xh - px) / Dx
if t1 > t2 then swap (t1, t2)
if t1 > tnear then tnear = t1
if t2 < tfar then tfar = t2
if tnear > tfar return FALSE
if tfar < 0 return FALSE
end

Return tnear
 Special Case
• Ray is parallel to an axis
– If Dx = 0 or Dy = 0 or Dz = 0

• px < x1 or px > x2 then miss

y =y2

y=Y1

D x=X1 x=X2
p
 Special Case
• Box is behind the eye
– If tfar < 0, box is behind
D
p

x = x1 x = x2