I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S

Andries van Dam September 23, 2008 ScanConversion2 #/26

Scan Conversion 2
(pages 81-91 in the textbook)
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26

Version 1: really bad
For x = R to R
y = sqrt(R * R x * x);
Pixel (round(x), round(y));
Pixel (round(x), round(-y));


Version 2: slightly less bad
For x = 0 to 360
Pixel (round (R cos(x)), round(R sin(x)));
(17, 0)
(0, 17)
(17, 0)
(0, 17)
Scan Converting Circles
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Symmetry: If (x
0
+ a, y
0
+ b) is on circle
also (x
0
a, y
0
b) and (x
0
b, y
0
a);
hence 8-way symmetry.

Reduce the problem to finding the pixels
for 1/8 of the circle



R
(x
0
+ a, y
0
+ b)
(x-x
0
)
2
+ (y-y
0
)
2
= R
2
(x
0
, y
0
)
Version 3 Use Symmetry
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Scan top right 1/8 of circle of radius R









Circle starts at (x
0
, y
0
+ R)
Lets use another incremental algorithm
with decision variable evaluated at
midpoint
(x
0
, y
0
)
Using the Symmetry
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
E
SE
x = x
0
; y = y
0
+ R; Pixel(x, y);
for (x = x
0
+1; (x x
0
) > (y y
0
); x++) {
if (decision_var < 0) {
/* move east */
update decision_var;
}
else {
/* move south east */
update decision_var;
y--;
}
Pixel(x, y);
}

Note: can replace all occurrences of x
0
, y
0
with 0, 0
and Pixel (x
0
+ x, y
0
+ y) with Pixel (x, y)
Shift coordinates by (-x
0
,-y
o
)
Sketch of Incremental Algorithm
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Decision variable
negative if we move E, positive if we
move SE (or vice versa).

Follow line strategy: Use implicit equation
of circle
f(x,y) = x
2
+ y
2
R
2
= 0

f(x,y) is zero on circle, negative inside,
positive outside

If we are at pixel (x, y)
examine (x + 1, y) and (x + 1, y 1)

Compute f at the midpoint
What we need for Incremental
Algorithm
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
P = (x
p
, y
p
)
M M
E
M
SE
SE
Decision Variable
E
Evaluate f(x,y) = x
2
+ y
2
R
2

at the point


We are asking: Is



positive or negative? (it is zero on circle)

If negative, midpoint inside circle, choose E
vertical distance to the circle is less at
(x + 1, y) than at (x + 1, y1).

If positive, opposite is true, choose SE
2
2
2
2
1
) 1 (
2
1
, 1 R y x y x f

|
.
|

\
|
+ + =
|
.
|

\
|
+
|
.
|

\
|
+
2
1
, 1 y x
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Decision based on vertical distance
Ok for lines, since d and d
vert
are
proportional
For circles, not true:








Which d is closer to zero? (i.e. which of the
two values below is closer to R):

R y x Circ y x d
R y x Circ y x d
+ + = +
+ + = +
2 2
2 2
) 1 ( ) 1 ( ) ), 1 , 1 ((
) 1 ( ) ), , 1 ((
2 2 2 2
) 1 ( ) 1 ( ) 1 (
+ + + +
y x or y x
The right decision variable?
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
We could ask instead:
Is (x + 1)
2
+ y
2
or (x + 1)
2
+ (y 1)
2

closer to R
2
?

The two values in equation above differ by
1 2 ] ) 1 ( ) 1 [( ] ) 1 [(
2 2 2 2
= + + + +
y y x y x
f
E
f
SE
= 290 257 = 33
(0, 17)
(1, 17)
(1, 16)
f
E
= 1
2
+ 17
2
= 290
E
SE
f
SE
= 1
2
+ 16
2
= 257
2y 1 = 2(17) 1 = 33
Alternate Phrasing (1/3)
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
The second value, which is always less,
is closer if its difference from R
2
is less than


i.e., if


then


so

so

so
<
2
) 1 2 (
1

y ] ) 1 ( ) 1 [(
2 2 2
+ +
y x R
) 1 2 (
2
1

|
.
|

\
|
y
2
1
) 1 ( 0
2 2 2
R y y x
+ + + <
2
1
1 2 ) 1 ( 0
2 2 2
R y y y x
+ + + + <
) 1 ( ) 1 (
2
1
0
2 2 2
R y x y
+ + + <
2
2
2
4
1
2
1
) 1 ( 0 R y x
+
|
.
|

\
|
+ + <
Alternate Phrasing (2/3)
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
The radial distance decision is whether



is positive or negative

And the vertical distance decision is whether



is positive or negative; d1 and d2 are
apart.


The integer d1 is positive only if d2 + is
positive (except special case where d2 = 0).




4
1
4
1
2
2
2
4
1
2
1
) 1 ( 1 R y x d
+
|
.
|

\
|
+ + =
2
2
2
2
1
) 1 ( 2 R y x d

|
.
|

\
|
+ + =
Alternate Phrasing (3/3)
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
How to compute the value of




at successive points?

Answer: Note that


is just


and that


is just

(1/2)
2
2
2
2
1
) 1 ( ) , ( R y x y x f

|
.
|

\
|
+ + =
2 2 3 2 ) , (
) , ( ) 1 , 1 (
3 2 ) , (
) , ( ) , 1 (
SE
E
+ + = A
+
+ = A
+
y x y x
y x f y x f
x y x
y x f y x f
Incremental Computation, Again
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
F
F
F
Incremental Computation (2/2)
If we move E, update by adding 2x + 3

If we move SE, update by adding
2x + 3 2y + 2.

Forward differences of a 1
st
degree polynomial
are constants and those of a 2
nd
degree
polynomial are 1
st
degree polynomials
this first order forward difference, like a
partial derivative, is one degree lower
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
The function is linear, hence
amenable to incremental computation, viz:







Similarly
Second Differences (1/2)
3 2 ) , (
E
+ = A
x y x
2 ) , ( ) 1 , 1 (
E E
= A + A
y x y x
2 ) , ( ) , 1 (
E E
= A + A
y x y x
4 ) , ( ) 1 , 1 (
SE SE
= A + A
y x y x
2 ) , ( ) , 1 (
SE SE
= A + A
y x y x
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
For any step, can compute new
E
(x, y) from old

E
(x, y) by adding appropriate second constant
increment update delta terms as we move.
This is also true of
SE
(x, y)

Having drawn pixel (a,b), decide location of new
pixel at (a + 1, b) or (a + 1, b 1), using
previously computed d(a, b).

Having drawn new pixel, must update d(a, b) for
next iteration; need to find either d(a + 1, b) or
d(a + 1, b 1) depending on pixel choice

Must add
E
(a, b)

or
SE
(a, b) to d(a, b)

So we
Look at d(i) to decide which to draw next,
update x and y
Update d using
E
(a,b) or
SE
(a,b)
Update each of
E
(a,b) and
SE
(a,b) for future
use
Draw pixel
Second Differences (2/2)
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26


Midpoint Eighth Circle Algorithm
MEC (R) /* 1/8
th
of a circle w/ radius R */
{
int x = 0, y = R;
int delta_E, delta_SE;
float decision;
delta_E = 2*x + 3;
delta_SE = 2(x-y) + 5;
decision = (x+1)*(x+1) + (y + 0.5)*(y + 0.5) R*R;
Pixel(x, y);
while( y > x ) {
if (decision > 0) {/* Move east */
decision += delta_E;
delta_E += 2; delta_SE += 2; /*Update delta*/
}
else {/* Move SE */
y--;
decision += delta_SE;
delta_E += 2; delta_SE += 4; /*Update delta*/
}
x++;
Pixel(x, y);
}
}
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Uses floats!
1 test, 3 or 4 additions per pixel
Initialization can be improved
Multiply everything by 4 No Floats!
Makes the components even, but sign of
decision variable remains same






Questions
Are we getting all pixels whose distance
from the circle is less than ?
Why is y > x the right stopping criterion?
What if it were an ellipse?

Analysis
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Patterned primitives



Aligned Ellipses



Non-integer
primitives



General conics
Other Scan Conversion Problems
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Patterned line from P to Q is not same
as patterned line from Q to P.





Patterns can be geometric or cosmetic
Cosmetic: Texture applied after
transformations
Geometric: Pattern subject to
transformations

Cosmetic patterned line


Geometric patterned line
Patterned Lines
P Q
P Q
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Geometric Pattern vs. Cosmetic Pattern
+
Geometric
(Perspectivized/Filtered)
Cosmetic
(Contact Paper)
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Equation is





i.e,




Computation of and is similar
Only 4-fold symmetry
When do we stop stepping horizontally and
switch to vertical?


Aligned Ellipses
1
2
2
2
2
= +
b
y
a
x
2 2 2 2 2 2
b a y a x b
= +
E
A
SE
A
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
When absolute value of slope of ellipse is
more than 1, viz:





How do you check this? At a point (x,y) for
which f(x,y) = 0, a vector perpendicular to
the level set is f(x,y) which is




This vector points more right than up when
Direction Changing Criterion (1/2)
(

(



c
c
c
c
) , ( ), , ( y x
y
f
y x
x
f
0 ) , ( ) , (
>
c
c

c
c
y x
y
f
y x
x
f
V
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
In our case,


and



so we check for



i.e.



This, too, can be computed incrementally
Direction Changing Criterion (2/2)
x a y x
x
f
2
2 ) , (
=
c
c
y b y x
y
f
2
2 ) , (
=
c
c
0 2 2
2 2
>
y b x a
0
2 2
>
y b x a
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26

Now in ENE octant, not ESE octant


This problem is artifact of aliasing much
more on this later
Problems with Aligned Ellipses
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Non-Integer Primitives
Initialization is harder
Endpoints are hard, too
making Line (P,Q) and Line (Q,R)
join properly is a good test
Symmetry is lost
Non Integer Primitives and
General Conics
General Conics
Very hard--the octant-changing test is
tougher, the difference computations
are tougher, etc.
do it only if you have to.

Note that drawing gray-scale conics is
easier than drawing B/W conics
I N T R O D U C T I O N T O C O M P U T E R G R A P H I C S
Andries van Dam September 23, 2008 ScanConversion2 #/26
Generic Polygons
(More information and these pictures on page
92-93 of textbook)