Dithering and Raster Graphics1

Dithering
1st color
missing colors are simulated by mixing existing colors
Dithering and Raster Graphics

Special Aspects of Raster Images
2nd color
missing color
2 /54
Eduard Grller, Thomas Theul
Dithering Methods (Digital Halftoning)

s
Dithering in Printing Industry

newspapers black ink on light paper, rasterization of the images enables also grey levels, equal point density everywhere, variable size q color printing every primary color is rasterized separately, different printing angles ensure unbiased results
q
threshold dithering x ordered dither x stochastic dither x dot diffusion x .... s error diffusion dithering (Floyd-Steinberg)
3 /54
4 /54
Threshold Dithering
every pixel is compared to a threshold t:

Constant Threshold Dithering

sample image threshold values result
t can be: q equal everywhere (e.g. (ba)/2, arbitrary value, mean value, median, ...) q location dependent (defined locally or globally)
Eduard Grller, Thomas Theul 5 /54
p<t p>t
a b
1 1 1 1
7 6 5 4
6 5 4 2
5 4 3 1
4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5 4.5
0 0 0 0
9 9 9 0
9 9 0 0
9 0 0 0
(values between 0 and 9) corresponds to rounding

Principle of Dithering
available values a, b missing value x between a and b shall be simulated by mixing a-pixels and b-pixels
Principle of Dithering (2)

a x-a x b
b-x b-a to produce color value x there have to be:
100 * x a % b-pixels ba 100 * b x % a-pixels ba

Eduard Grller, Thomas Theul 7 /54 Eduard Grller, Thomas Theul 8 /54
Dithering a Uniform Area

for a uniform area regular application of this pattern will produce 1/4 1/2 3/4 this grey tone 0
Dithering a Uniform Area (2)

This can be done by using a different threshold for every pixel (using the interval borders)
1/8
5/8 3/8 5/8 ...
1/8 7/8 1/8
5/8 3/8
1/8 7/8
5/8 ...
1
7/8
interval borders
1/8
3/8
5/8
7/8
1/8
= "Threshold Matrix"
all grey levels in this interval will be mapped to 1/4

9 /54
10 /54
Threshold Matrix
distances between interval borders are equal, therefore it suffices to define the sequence of the pixel values in the matrix:
Dither Matrix Example

dither matrix threshold matrix
7 0 5
2 4 6
3 8 1
1/8
5/8
2 1
instead of
7/8 3/8
only
3
15 18 1 18 11 18
5 7 18 18 9 17 18 18 13 3 18 18
2k+1 2n2
12 /54
i.e. for an nxn matrix: values [0,n21]

value k corresponds to threshold value:
2k+1 2n2
11 /54
value k corresponds to threshold value:

Threshold Matrix Dithering Example

sample image threshold values result
Generation of Threshold Matrices (1)

recursive method: 4 copies of smaller matrices
1 1 1 1
7 6 5 4
6 5 4 2
5 4 3 1
1.1 5.6 1.1 5.6 7.9 3.4 7.9 3.4 1.1 5.6 1.1 5.6 7.9 3.4 7.9 3.4
0 0 0 0
9 9
0
9
0
9
0 0
0 2 3 1
...
k= 0(1)n2-1
4k+0
4k+2
9
0
(values between 0 and 9)

example: 0 12 3 15
8 4 11 7
2 14 1 13
10 6 9 5
4k+3
4k+1
14 /54
Generation of Threshold Matrices (2)

direct method: use of magic squares
Dithering between Grey Levels

a x b
example
magic squares produce fewer diagonal stripes

calculation is done separately for every pixel (not once for a dithering matrix)
Grey Level Dithering Example

4 grey values are available: 0, 3, 6, 9 0 dither matrix: 3
Dot Diffusion Dithering

2 1
ordering of the threshold values generates larger dot areas example:
sample image threshold values
result
1 1 1 1
7 6 5 4
6 5 4 2
5 4 3 1
0.4 7.9 3.4 4.9 2.6 3.4 5.6 4.1 0.4 4.9 3.4 1.9 2.6 4.1 2.6 1.1
6 6 6 3
6 3 6
6 3 3
0
3
11 16 12 8 4 5 15 17 13 3 0 1 10 14 9 7 2 6
simulates traditional printing techniques for high resolution devices
0
17 /54
(values between 0 and 9)
0 11 12 7
14 3 5 8 2 15 9 4
13 6 1 10
threshold values have to lie between a and b:
a + 2k+1 (b-a) 2n2
Stochastic Dithering?
use of random numbers as threshold values
s expectation s no
Forced Random Matrix Dithering

very good results improved "random" matrices method: insert threshold values one by one into matrix, always use the position farthest away from all previous points r p repulsive force field: f(r) = e s precalculate large threshold matrices: 300x300 very good results!

value of total error = 0 regular artificial patterns possible
unfortunately: very bad results! (due to bad distribution of random numbers)
19 /54
Error Diffusion Dithering

(Floyd-Steinberg)
the rounding error of every pixel is propagated to neighbor pixels and compensated there variations: which neighbor pixels are affected?
Simplest Error Diffusion Dithering

pixel line correct value rounded value ... k1 k2 k3 k4 k5 ... r1 r2 r3 r4 r5 ...
r1 := round (k1) error1 := r1 k1 r2 := round (k2 error1) error2 := r2 (k2 error1) ... ri := round (ki errori1) errori := ri (ki errori1)
21 /54 Eduard Grller, Thomas Theul 22 /54
Error Diffusion Dithering Example

example: the values 0, 3, 6, 9 are available
Diffusion Direction Variations

to gain better results, the error is distributed to several neighbors (with weights) sum of all weights = 1 1 x x 7 16 3 5 1 often used x 7 1 x weighting 48 3 5 7 5 distributions: 1 x 1 1 3 5 3 x 2 1 x 8 1 x 42 2 4 8 4 1 2 4 2
k: 1 1 1 2 3 4 7 1 5 ... r: 0 3 0 3 3 3 6 3 3 ... f: -1 1 0 1 1 0 -1 1 -1 ... ri := round (ki - fi-1) fi := ri - ki + fi-1
5 3 1 4 2 1
23 /54
24 /54
Error Diffusion Dithering Example

example: the values 0, 3, 6, 9 are available 1 error distribution: 2 sample image error
-1 1.5 .75 1.37
x
Serpentine Method
artificial stripes can be reduced drastically by processing the scanlines in serpentine order instead of "normal" now in "serpentines"
x 1 1
result
1 1 1 1
7 6 5 4
6 5 4 2
5 4 3 1
0
3
9 6 3 6
6 3 6
6 3 3 3
25 /54
1.5 1.5 -.87 -.75 -.25 -1.37 .87 .06 -1.12 .75 -1.12 1.47
0 0
...
no additional memory necessary
...
(values between 0 and 9) Eduard Grller, Thomas Theul
26 /54
Raster Conversion
converting primitives (lines etc.) into pixels very frequent operations, therefore necessary: s efficiency s possibility to implement in hardware
Raster Conversion of Lines

lines should appear straight lines should appear uniformly bright lightness should be independent of direction endpoints should be "exact"
27 /54
28 /54
Symm. DDA for Line (x1,y1) (x2,y2)

(x,y) := ( (x2,y2) (x1,y1) ) / 2i draw (rd(x1) / rd(y1)) (x1,y1) := (x1+x,y1+y) x2 < x1+x YES NO rd(x1x) rd(x1) ORrd(y1y) rd(y1) YES
Simple DDA
symmetric DDA produces lines of variable breadth, but if x and y are chosen such that max(|x|, |y|) = 1 only 1 pixel per unit in the longer direction

2i < length i-1 <2
draw (rd(x2) / rd(y2))
problem: requires a real division solution: Bresenham Algorithm image equal to simple DDA, but no division
NO
STOP
Comparision Symmetric - Simple DDA
Raster Conversion of Lines

(Bresenham's Line Algorithm)
T D H
} dy yD-yS < yT-yD yD-yS > yT-yD
S T
symmetric DDA
simple DDA
31 /54
only for lines with angle 45 other lines by mirroring/rotating with 90/180...
Some Bresenham Mathematics

yD yS < yT yD yD yH < 0 yD yS > yT yD yD yH > 0
Bresenham's Line Algorithm (1)

ys:=y1; d:=-0.5; {d=yDyH} dy:=(y2-y1)/(x2-x1); FOR xs:=x1 TO x2 DO BEGIN SetPixel(xs,ys); d:=d+dy; IF d>0 THEN BEGIN ys:=ys+1; d:=d-1 {because yH:=yH+1} END END
in every case yD := yD + dy and (xS/yS) are drawn let d = yD yH

Bresenham's Line Algorithm (2)

ys:=y1; e:=-(x2-x1)div 2; {d*(x2-x1)} de:=(y2-y1); {dy*(x2-x1)} FOR xs:=x1 TO x2 DO BEGIN SetPixel(xs,ys); e:=e+de; IF e>0 THEN BEGIN ys:=ys+1; e:=e-(x2-x1) END only integers! END only addition, subtraction, shift!
if yD yH < 0 if yD yH > 0
yS, yT do not change, yS := yS+1, yT := yT+1
decision variable
Raster Conversion of Circles

(Bresenham's Circle Algorithm)
utilize threefold symmetry!
only one eighth has to be calculated
36 /54
Bresenham's Circle Algorithm

P T H S
yD-yS > yT-yD yD-yS < yT-yD


yDyS > yTyD H inside the circle that is if xH2+yH2 < r2 or if f(xH,yH) = xH2+yH2 r2 < 0

(only for 2nd octant)


if dold < 0 then Hnew = Hold + (1,0) , ie. dnew = f(xp+2, yp1/2) = (xp+2)2 + (yp1/2)2 r2 dnew = dold + (2xp+3) if dold > 0 then Hnew = Hold + (1,1) , ie. dnew = f(xp+2, yp3/2) = (xp+2)2 + (yp3/2)2 r2 dnew = dold + (2xp 2yp + 5)

initialization: restriction to integer radii
x=0

and
y=r
H = (1,r-1/2) d = f(H) = 1 + (r2-r+1/4) - r2 = 5 /4 - r
T S
d = "decision variable" i.e. dold = d = f(xp+1, yp1/2) , then dnew can be calculated as follows:

PROCEDURE Circle (r:INTEGER); VAR x,y: INTEGER; d: REAL; BEGIN ... {initialize x,y,d} REPEAT Draw_8(x,y) IF d < 0 THEN d:=d+2*x+3 ELSE BEGIN d:=d+2*(x-y)+5; y:=y-1 END; x:=x+1; UNTIL y < x Eduard Grller, Thomas Theul END;

PROCEDURE Circle (r:INTEGER); VAR x,y: INTEGER; d: REAL; BEGIN x:=0; y:=r; d:=5/4 - r; REPEAT Draw_8(x,y) IF d < 0 THEN d:=d+2*x+3 ELSE BEGIN d:=d+2*(x-y)+5; y:=y-1 END; x:=x+1; UNTIL y < x Eduard Grller, Thomas Theul END;
41 /54
criterion:
f(xH,yH) < 0 f(xH,yH) > 0
T S
40 /54
42 /54
Raster Transformations
how to apply geometrical transformations to raster images? translation: trivial scaling: resampling necessary s shearing: by line conversion s rotation: partition in three shearings
s s
Raster Scaling: Scaling Up

old resolution new resolution
center point of pixel in new resolution defines its color = "resampling"

Raster Scaling: Scaling Down

old resolution new resolution
Shearing
x-shearing
(x y) 1 0 = (x+y y) 1
y-shearing
(x y) 1 = (x y+x) 0 1
center point of pixel in new resolution defines its color = "resampling"

46 /54
Raster Shearing
Raster Rotation
90-rotation, 180-rotation, etc. trivial
principle of other rotations: subdivide rotation in a series of three shears
can be seen as the multiple application of a line raster conversion algorithm (e.g. Bresenham) no information lost

cos sin -sin cos
1 0 . 1 . 1 0 1 0 1 1
where = - tan (/2) and = sin

47 /54
Rotation = 3 Shears
tan (/2) = (1- cos ) / sin 1 cos 1 sin 0. 1 1 sin 0 1 . 1 0 cos 1 1 = sin
Example: 45 Rotation
1 cos 1 sin
0 1 sin . 1 1 = cos cos sin
tan (/2) = 0.4142 sin = 0.7071
sin 1+ cos 1 cos sin cos 1+ cos2 cos cos = -sin cos sin
Eduard Grller, Thomas Theul 49 /54 Eduard Grller, Thomas Theul 50 /54
tan (/2) = 0.4142 sin = 0.7071
tan (/2) = 0.4142 sin = 0.7071
51 /54
52 /54
Distortion of Raster Images

v y
Distortion of Raster Images

y u=0 interpola tion interpola tion u v=0 v=1 interx pola tion (u,v) v u=1 u=1 u=0
we are looking for the (u,v)-coordinates of the center of pixel (x,y)

(u,v) are calculated from (x,y) by interpolation from the corner coordinates

Dithering and Raster Graphics1

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Dithering and Raster Graphics1

Hochgeladen von

Copyright:

Verfügbare Formate

Dithering

Dithering and Raster Graphics

Eduard Grller, Thomas Theul

Dithering Methods (Digital Halftoning)

Dithering in Printing Industry

Eduard Grller, Thomas Theul

Eduard Grller, Thomas Theul

Constant Threshold Dithering

(values between 0 and 9) corresponds to rounding

Principle of Dithering (2)

b-x b-a to produce color value x there have to be:

100 * x a % b-pixels ba 100 * b x % a-pixels ba

Dithering a Uniform Area

Dithering a Uniform Area (2)

5/8 3/8 5/8 ...

1/8 7/8 1/8

Eduard Grller, Thomas Theul

all grey levels in this interval will be mapped to 1/4

Eduard Grller, Thomas Theul

Dither Matrix Example

i.e. for an nxn matrix: values [0,n21]

value k corresponds to threshold value:

Threshold Matrix Dithering Example

Generation of Threshold Matrices (1)

(values between 0 and 9)

Generation of Threshold Matrices (2)

Dithering between Grey Levels

magic squares produce fewer diagonal stripes

Grey Level Dithering Example

Dot Diffusion Dithering

ordering of the threshold values generates larger dot areas example:

sample image threshold values

Eduard Grller, Thomas Theul

(values between 0 and 9)

threshold values have to lie between a and b:

a + 2k+1 (b-a) 2n2

Forced Random Matrix Dithering

value of total error = 0 regular artificial patterns possible

unfortunately: very bad results! (due to bad distribution of random numbers)

Eduard Grller, Thomas Theul

Error Diffusion Dithering

Simplest Error Diffusion Dithering

Eduard Grller, Thomas Theul

Error Diffusion Dithering Example

Diffusion Direction Variations

k: 1 1 1 2 3 4 7 1 5 ... r: 0 3 0 3 3 3 6 3 3 ... f: -1 1 0 1 1 0 -1 1 -1 ... ri := round (ki - fi-1) fi := ri - ki + fi-1

Eduard Grller, Thomas Theul

Error Diffusion Dithering Example

(values between 0 and 9) Eduard Grller, Thomas Theul

Raster Conversion of Lines

Eduard Grller, Thomas Theul

Eduard Grller, Thomas Theul

Symm. DDA for Line (x1,y1) (x2,y2)

2i < length i-1 <2

draw (rd(x2) / rd(y2))

Comparision Symmetric - Simple DDA

Raster Conversion of Lines

} dy yD-yS < yT-yD yD-yS > yT-yD

Some Bresenham Mathematics

Bresenham's Line Algorithm (1)

in every case yD := yD + dy and (xS/yS) are drawn let d = yD yH

Bresenham's Line Algorithm (2)

yS, yT do not change, yS := yS+1, yT := yT+1