CS 430/536

Computer Graphics I

Thick Primitives,
Halftone Approximation
Anti-aliasing
Week 6, Lecture 11

David Breen, William Regli and Maxim Peysakhov

Geometric and Intelligent Computing Laboratory
Department of Computer Science
Drexel University
1
http://gicl.cs.drexel.edu
 Outline
•  Drawing with Thick Primitives
•  Halftone Approximation
•  Anti-aliasing

2
 Drawing with Thick Primitives
•  How do we thicken
the line stroke width?
•  Ideas:
–  Place the center of
the circular “brush” on
the pixel
–  Place the upper
corner of the square
“marker” on the pixel
(issues of orientation)
–  Then do scan
conversion algorithm 3
 Three Basic Methods
1.  Column Replication
–  Use >1 pixel per col/row
2.  Trace brush outline
across 1-pixel primitive
3.  Trace two copies, t
apart, and fill in

4
1994 Foley/VanDam/Finer/Huges/Phillips ICG
 Column(Row) Replication
•  Idea: duplicate pixels in
–  Columns, when
-1 < slope < 1
–  Rows, otherwise
•  Thickness t is from
primitive’s boundaries
perpendicular to its tangent
•  What happens for even t?
•  Issues when lines meet at
angles, when octants
merge, brightness for
sloped lines, etc. 5
1994 Foley/VanDam/Finer/Huges/Phillips ICG
 Moving the Pen
•  Example:
–  a rectangular pen
–  Center or corner follows
scan algorithm
•  How to implement?
–  Idea 1: fill the box at each
point
–  Problem: pixels get colored
more than once
–  Idea 2: fill by using a span of
the pen primitive at each
step
6
1994 Foley/VanDam/Finer/Huges/Phillips ICG
 Halftone Approximation
•  Not all devices can display all colors
–  e.g. GIF is only 256 colors
•  Idea: With few available shades,
produce illusion of many colors/shades?
•  Technique: Halftone Approximation
•  Example: How do we do greyscale with
black-and-white monitors?
7
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation
•  Technique: Dithering
•  Idea: create meta-pixels, grouping base
pixels into 3x3s or 4x4s
•  Example: a 2x2 dither matrix for grayscale

8
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation
•  Issues with Dithering
–  Image is now 4x in size
•  How do we keep image the same size?
•  Technique: Error Diffusion
•  Idea: When approximating pixel
intensity, keep track of error and try to
make up for errors with later pixels
9
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation:
Error Diffusion Example #1
•  Problem: draw 1D line with 1/3 gray tone
•  Pixel #1: round to black, 0… error 1/3
•  Pixel #2: value 1/3+1/3=2/3, color white
•  Pixel #3: value 1/3-1/3=0, color black
•  Pixel #4: value 1/3+0= 1/3, color black
•  Color sequence: 01001001001…

10
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation:
Error Diffusion Example #1
Draw 1/3 gray line

•  Pixel: 1/3 1/3 1/3 1/3 1/3 1/3 1/3 1/3

•  Error: 0 1/3 -1/3 0 1/3 -1/3 0 1/3
•  FB: 0 1 0 0 1 0 0 1

•  Color sequence: 01001001001…

11
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation:
Error Diffusion Example #2
•  Consider a 2D image/primitive
–  Goals: Spread errors out in x and y pixels
Nearby gets more error than far away
–  Floyd-Steinberg Error Distribution Method
•  Let the current pixel be
•  Distribute error as follows:

12
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation:
Error Diffusion Example #2
•  Let be the shade of pixel
•  To draw we round pixel to nearest
shade and set
•  Then, diffuse the errors throughout
surrounding pixels, e.g.
S[x + 1][y] += (7/16)  err
S[x - 1][y - 1] += (3/16)  err
S[x][y - 1] += (5/16)  err
S[x + 1][y - 1] += (1/16)  err 13
Pics/Math courtesy of Dave Mount @ UMD-CP
 Halftone Approximation:
Error Diffusion Example #2

14
Pics courtesy of Dimitri Gusev @ Indiana
 Antialiasing
•  Converting an idealized line to a discrete grid
is an approximation
•  “Jaggies” are the result

15
1994 Foley/VanDam/Finer/Huges/Phillips ICG
 The Aliasing Problem
•  General problem in
Analog-to-Digital
conversion
–  When sampling, one
needs to sample at a
higher frequency
than the analog
signal
Example: Music CDs
–  Aliasing shows up as Sampled: 44Khz
spurious low Max frequency: 22Khz
frequencies 16
1993 ACM SIGGRAPH Education Slide Set
 The Aliasing Problem
•  General problem in
Analog-to-Digital
conversion
–  When sampling, one
needs to sample at a
higher frequency
than the analog
signal
Example: Music CDs
–  Aliasing shows up as Sampled: 44Khz
spurious low Max frequency: 22Khz
frequencies 17
1993 ACM SIGGRAPH Education Slide Set
 Zone plate
•  Zone plate
–  f(x,y) = x2 + y2
–  Black if floor(f) odd, else
white
•  Outer rings occur too often to
be sampled correctly
•  Moiré patterns resemble the
zone plate

18
Compiled
from:
Lecture
notes
of
Dr.
John
C.
Hart
@
University
of
Illinois
 Aliasing in Computer Graphics
•  Mathematical model
of image: analog
•  Screen: digital
•  Result: visual
effects, jaggies, lost
textures and detail

19
1993 ACM SIGGRAPH Education Slide Set
Antialiasing in use…

20
 Antialiasing
•  How to create the
visual effect of Two major categories of
smoothing the line? anti-aliasing techniques:
(1) PreFiltering
•  We need to find a (2) PostFiltering
way to simulate the
display of partial
pixels

21
1994 Foley/VanDam/Finer/Huges/Phillips ICG
 Antialiasing: PreFiltering
•  Idea:
–  each pixel has area
–  compute color based
on overlap with the
object’s area
•  Result: smoother
borders for
objects

22
1993 ACM SIGGRAPH Education Slide Set
 Antialiasing: PostFiltering
(SuperSampling)

•  Idea:
–  take multiple samples for each pixel
–  create weighted measure of color
–  samples can be stochastic or regular
–  stochastic makes for more pleasing results 23
1993 ACM SIGGRAPH Education Slide Set
 Antialiasing: PostFiltering
•  Filters: the weighted measure of color
–  weighted or unweighted average of the pixels

24
1993 ACM SIGGRAPH Education Slide Set
Antialiasing of Lines & Curves
•  Example: draw 1-pixel thick black line
2 different screen resolutions =>
2x jags, jags 1/2 size, but 4x memory!

25
1994 Foley/VanDam/Finer/Huges/Phillips ICG
Antialiasing of Lines & Curves

•  Ideally, the line should be 1 pixel wide

–  If line is vertical or horizontal, there are no issues
•  Otherwise, there are issues. The basic idea:
–  Represent line l as a narrow rectangle Rl
–  Represent pixels as non-overlapping square points
–  Assign pixel intensity based on portion of pixel Rl covers
•  Totally covered = 1.0
•  Partially covered = 0.25 --- 0.50 --- 0.75
•  Not covered = 0.0
26
1994 Foley/VanDam/Finer/Huges/Phillips ICG
Weighted Area Sampling:
Box Filters
•  Pixel intensity
Imax = 1, Imin = 0
•  Box over the pixel:
weighting fctn
•  The volume Ws is the
weight
–  Ws * Imax
•  (achieves unweighted
area sampling)
27
1994 Foley/VanDam/Finer/Huges/Phillips ICG
Weighted Area Sampling:
Cone Filters
•  Idea:
–  Cone centered at each pixel
–  Radius of cone bigger than
single pixel
–  Volume of cone is 1
–  Intensity is volume Ws
•  Achieves
–  Linear decrease of intensity
vs distance
–  Smoother effects by
decreasing pixel contrasts
–  Rotational symmetry
28
1994 Foley/VanDam/Finer/Huges/Phillips ICG
 Antialiasing Area Boundaries
Supersampling
•  Double (triple,quadruple) the resolution of your buffer.
•  Use any polygon filling algorithm to scan-fill the polygon
in BW
•  Compute the color of each pixel by counting the
number of colored sub-pixels

29
Anti-aliasing of Circles and Curves
•  Super-sampling can be applied to anti-aliasing of circles
and curves
•  Double (triple,quadruple) the resolution of your buffer.
•  Draw the circle of the curve with double
(triple,quadruple) width
•  Compute the color of each pixel by counting the
number of colored sub-pixels

30
Antialiasing with Grayscales

31
From HW of Adrian Burka
Antialiasing with Grayscales

32
From HW of Adrian Burka
 Anti-aliasing from CG II

Steven Thomas 33
 Antialiasing Example

34
1993 ACM SIGGRAPH Education Slide Set