You are on page 1of 16

This article was downloaded by: [University of Windsor]

On: 19 November 2014, At: 06:02


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Journal of Mathematics and the Arts


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tmaa20

On fine art and generative line drawings


H. Dehlinger

Stiegelwiesen 3 , D-34132 Kassel, Germany


Published online: 06 Aug 2007.

To cite this article: H. Dehlinger (2007) On fine art and generative line drawings, Journal of Mathematics and the Arts, 1:2,
97-111, DOI: 10.1080/17513470701441445
To link to this article: http://dx.doi.org/10.1080/17513470701441445

PLEASE SCROLL DOWN FOR ARTICLE


Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

Journal of Mathematics and the Arts,


Vol. 1, No. 2, June 2007, 97111

On fine art and generative line drawings


H. DEHLINGER*
Stiegelwiesen 3, D-34132 Kassel, Germany
(Received 8 January 2007; in final form 8 May 2007)

Downloaded by [University of Windsor] at 06:02 19 November 2014

The algorithmic generation of line drawings is discussed from an artists point of view.zUsing artwork examples
generated by the author, several different methods are considered and the underlying logic and algorithms are
presented. A descriptive model for line drawings is proposed, viewing the universe of hand drawings and the universe
of machine drawings as distinct entities. This model allows us to consider the range of possibilities for line-oriented,
algorithmically generated drawings.
Keywords: Generative art; Algorithmic art; Line drawing; Fine art
AMS Subject Classification: 00A72

1. Introduction
The point of departure for the work described here is
the conception, the programming, and the execution of
fine art line drawings generated using algorithms. The
drawings considered are somewhat limited in scope in
the sense that they are strictly straight line oriented,
executed on pen-plotters, and drawn on paper using
special pens. The programs for such drawings rely on
definitions together with generative rules that are viewed
as manifestations of the intentions of the artist. Since
they are purposefully conceived and designed before
the drawings are executed, the design and coding of the
generative rules is an integral part of the work. They
brand-mark the art pieces in a very distinct way.
The history of art that relies on the use of computers
is short yet rich in its diversity of approaches. Even if
we restrict our attention to line-oriented work only, an
astonishing variety of approaches become apparent. Art
works by Manfred Mohr [1] (especially the works from
1969 to 1973), Vera Molnar [2], Frieder Nake [3], Georg
Nees [4], Mark Wilson [5], Harold Cohen [6], Charles
Csuri [7], Jean Pierre Hebert [8], Robert Krawczyk
[9, 10], Roman Verostko [11], and the early images
published by the Tokyo Computer Technique Group
[12] all reveal distinctive and very unique ways to
generate drawings with the aid of computer programs.
However, the core concept underlying all of them, which
is to rely on some kind of premeditated rule system as
a key element of an art production process, is found
*Email: dehling@uni-kassel.de

throughout art history. In a wider sense, the striking


homogeneity of a-perspective Egyptian art, the column
order of Greek and Roman architecture, the pattern
language of Islamic tiling, the brush strokes and line
types used by van Gogh, the rules used in the art of
etching before the dawn of photography, certain aspects
of the serial variations in the Homage to the Square by
Albers [13], the hypothesized self-similarities in Pollocks
drip paintings [14], and the rule based, line oriented
work by Sol Le Witt [15] are but a few of the many
examples where rules can be identified as principal
components that help shape the art. We might therefore
justifiably say that any body of work with a recognizable
identity arises from applying some set of rules that the
artist has conceived or designed. Such an artists set of
procedures might be compared to a musical score that
embodies the individual style of a composer. From this
point of view, generative art and algorithmic art created
using computers are but a variation of a theme long
known to art in general. There is one important
distinction, however, from earlier times: the strictness
of keeping to the rules and the precision in their
execution demanded by the computer are unparalleled.
Accepting the computer as part of the art making
equation, the rule sets artists devise now have to bow
without mercy to this new order. As a reward for
accepting this state of affairs, artists are granted the
privilege to explore totally new and hitherto unknown
domains.
In this paper I demonstrate the surprisingly rich
variation in generative art drawings that can occur,

Journal of Mathematics and the Arts


ISSN 17513472 print/ISSN 17513480 online 2007 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/17513470701441445

98

H. Dehlinger

despite the severe restrictions I impose upon them, by


giving several examples of my drawings. My primary
focus is on the generative (i.e., design and creation) side
of the problem of executing line drawings. I make no
attempt to apply computational measures of quality to
my drawings or to automate in any way the aesthetic
judgment of my drawings. However, I do attempt to
frame the discussion within the context of what I call the
Universe of Machine Generated Drawings.

Downloaded by [University of Windsor] at 06:02 19 November 2014

2. Methodology
The generative processes I use for my drawings can best
be illustrated through examples. Consider the drawings
in figures 1 and 2. Although they are clearly different
from one another, they were selected to illustrate two
extremes lying on a continuum of designs for generative
processes that I have explored. I will now consider each
in turn.

Figure 1.

2.1. Examples of generative drawings


The artistic concept behind the drawing in figure 1 is the
use of a single drawing element to generate the entire
drawing. In figure 1 this element is a short horizontal
stroke of constant length. Although this is a rather
simple idea, the key feature that underlies this generative
process is the repetition of the drawing element n times
where n is very large (n4 4103), coupled with the
enforcement of strict rules governing the relative
positions where the drawing elements can be placed.
I have executed a series of drawings using this concept
by choosing different drawing elements such as circles,
triangles, or short poly-lines. The inspiration for this
series of drawings arose from thinking about what one
sees when one is looking at a large freshly mown lawn
or when one is walking along a pebble beach. The
generative rules that were programmed for these
drawings are concerned with arrangements of the
drawing element in some regular fashion using a
very large number of identical, or nearly identical,

T_x_01 (top left corner), 2005. Generative line drawing, pen and plotter. Copyright H. Dehlinger.

99

Generative line drawings

into programs. To summarize, the essential features of


the process used to generate the drawing in figure 1 are:
. A short, horizontal stroke of constant length is drawn
n times to form a vertical strip.
. The distance between successive strokes within a strip
varies and is determined by a pseudo random number
chosen to lie within a specified range.
. There are m vertical strips generated, and these strips
are arranged parallel and flush with one another.
The drawing in figure 2 relies on an entirely different
artistic concept for its generative construction. The
concept behind this drawing is best described as digital
collage. Traditionally, a collage is an artistic composition in which different materials are glued to the canvas.

Downloaded by [University of Windsor] at 06:02 19 November 2014

elements [16]. I believe, despite its simplicity, that this is


a rather philosophical theme. In the case of figure 1, a
uniform arrangement was used. This meant that there
was only one variable to control, the spacing between
the strokes. My point is that to gain greater artistic
control over this somewhat general theme it had to be
transformed into a precisely defined set of rules.
When designing any generative process, at the outset
many possibilities suggest themselves. To sort through
these, I typically begin by performing a number of
experiments. At first, I test rather simple ideas, in part
because they are the easiest to program, but also because
the results of simple experiments often reveal a lot about
their future potential. These experiments provide guidance for encoding the rule sets under consideration

Figure 2.

b204.ar.plt_03, 2005. Generative line drawing, pen and plotter. Copyright H. Dehlinger.

Downloaded by [University of Windsor] at 06:02 19 November 2014

100

In my digital collages for my materials I use poly-lines


as drawing elements and polygonal shapes as clipping
shapes. Clipping is a well-known method for precisely
trimming and cutting lines and it is widely used in
software for computer aided design (CAD) and mechanical drawings. There are a number of clipping algorithms
to choose from. For figure 2, I chose a variation of the
commonly used Sutherland-Hodgman clipping algorithm [17]. The effect of clipping applied to a drawing
can be twofold: either it cuts away a piece of the drawing
leaving behind a blank space in the shape of the clipping
polygon, or it removes all lines from the drawing except
those that lie inside the clipping polygon. In the latter
case, the drawing is thereby reduced in size and trimmed
to precisely fit that shape. The materials used for gluing
the collage shown in figure 2 are digital in nature and
were all generated based on pseudo random numbers
lying within specified ranges. The digital collage
elements that are my materials are applied sequentially
in pseudo random order. I will now give a description of
a hypothetical collage scenario using my method. First a
large number of poly-lines is generated and cast onto the
plane. In subsequent generative steps more sets of such
poly-lines are generated and a decision is made about
whether or not they will be added to the drawing.
During each of these generative steps values for the
variables controlling the placement of the poly-lines
within regions, the total number of poly-lines, and the
attributes of the poly-lines must also be provided.
At certain random points in the generative sequence,
the application of clipping shapes will occur. It is
important to note that by using an interactive artistic
evaluation decision mechanism, the contribution from
each discrete sequence of poly-line generation followed
by polygon clipping to the cumulative result may be
either kept or discarded. Concerning clipping itself,
recall that in CAD-systems its sole use is for trimming
lines. In contrast, here since clipping is part of an artistic
process, clipping may sometimes become fuzzy clipping
[18]. This means that if n lines are presented to the fuzzy
clipping algorithm then roughly m lines will escape being
clipped, where m depends on the degree of fuzziness
desired. This technique enhances the generative method
just described to provide more interesting drawings.
Returning once again to figure 2, in more poetic terms,
we might say that poly-lines of different lengths are
floating across the plane and extending in many
directions. We might notice zones of dense lines and
zones where clipping-shapes are clearly visible. We
might see lines extend into the empty spaces of the
plane, delicate and thin in nature but prominently
visible.
The two examples discussed in this section have
indicated two extremes of drawings that can be obtained
using generative methods. Between these two extremes

H. Dehlinger

Figure 3. tree_11, 2001. An example of generative minimalism. Generative line drawing, pen and plotter. Copyright H.
Dehlinger. See insert for colour version of this figure.

lies a wide spectrum of possibilities. To further


complicate matters, drawings can be generated as onepass drawings, or multi-pass drawings using a number
of sequential steps. Another example of a drawing, using
the same generative algorithms as figure 2 but resulting
from a one-pass generative process, is given in figure 3.
When drawings are generated as one-pass drawings,
no visual feedback or loop control variable is available
to the artist. In this case the drawings concept is
completely embedded in the generative rules. This
production process differs significantly from traditional
art making processes where feedback is an essential
quality for both developing and completing the work.
I am convinced that the absence of such feedback
when using a generative approach contributes unique
characteristics to the resulting imagery. As an artistic
challenge it is interesting to go even further when
developing concepts for generative drawings by
demanding, for example, that the generative rules use
a minimal set of drawing commands. In this case one
could say that an artist is engaging in generative
minimalism [19]. The challenge of overcoming the
resistance of the material arises frequently in the
context of art making, and it finds its computational
pendant in such considerations. All of the examples in
this paper are due to algorithms and programs written
by me, but it should be noted that in addition to relying
on higher level programming languages, standard software (e.g. CAD programs, plotting programs, drawing

101

Generative line drawings

programs) and specific library routines available in


standard software packages were also involved in the
design of the generative processes.
2.2. An algorithm for poly-line generation

Downloaded by [University of Windsor] at 06:02 19 November 2014

The following algorithm for generating lines is intended


to be mathematically precise. It addresses the case where
line generation occurs in two dimensions only.
The algorithm yields points pk for k 1, 2, . . . , N
where N441. These points are organized into polylines P0, P1, . . . , PM where the jth poly-line Pj is
generated as follows:
1. The starting point Q0 of poly-line Pj is pseudo
randomly chosen to lie in some region R0;
2. The number nj of segments for poly-line Pj is pseudo
randomly chosen to lie in the set {m, m 1,
m 2, . . . , m n}, where the integers m40, n40
are specified by the artist;
3. The successive deviation angles j(0), j(1), . . . ,
j(nj  1), for the line segments are pseudo randomly
chosen so that j(0) is the angle for the first segment
of the poly-line Pj, and the angles j(1), . . . , j(nj  1)
satisfy the constraint that for s 0, 1, . . . , nj  2,
j(s) j (s  1)  j (s), lies in the interval [, ],
where 40 is specified by the artist;
4. The angular boundary  for a poly-line lies in the
interval [1, 2] where 0 51525360 ;
5. The segment lengths j(0), j(1), . . . , j(nj  1) of the
poly-line Pj are pseudo randomly chosen so that they
lie in the interval [a, b] where b4a40, and a and b
are specified by the artist.
The drawing loop for the generative process is then
for j 1, 2, . . . , M draw poly-line Pj.
The above algorithm can be easily generalized to
three dimensions although I have not pursued this.
However, a two-and-a-half dimensional approach has
been considered and some very interesting results were
obtained (see, for example, figure 4).
Some additional remarks may be helpful to further
highlight some of the ideas introduced in the linegenerating algorithm. As we have seen, most of the
drawings discussed here are based on poly-lines. This
is a deliberate choice. These lines are described by a
sequence of features that can be controlled using
parameter values that lie in intervals. Features that can
be incorporated into the line generation algorithm
include: the maximal number of segments in a polyline (e.g. to 2000), the angular boundary for a poly-line
(e.g. 0 to 360 ), and the angle deviation of a segment
from the preceding segment (e.g. 0 to 360 ). Moreover,
any segment of a poly-line can be required to have a
fixed length or a fixed angle or both. The stem of the
tree in figure 3 was generated by requiring the first

Figure 4. Pile. 2006. Generative line drawing, pen and


plotter. Copyright H. Dehlinger.

stroke of every poly-line generated to be drawn with


an angle of 90 .
Besides algorithmically controlled features, in my
drawings mechanical features of the plotter must also be
considered. These include pressure on the pen (e.g. 15 g
to 450 g), speed of the pen (e.g. 10 cm/sec to 80 cm/sec),
and acceleration of the pen (e.g. maximum 3.0 G).
Finally, during the execution of a drawing, the type of
pen (pencil, ink, ballpoint etc.), the thickness of the
plotted line, and the type of paper used must be also
considered. Besides the attributes of the lines themselves,
for each drawing executed, parameters such as the
number of starting points to use for lines (e.g. up to 2000
per drawing) and the number of lines that may originate
from a given point (e.g. up to 2000 per drawing) have
to be specified. For me, the distribution pattern of
starting points for poly-lines is another important
setting. Possibilities for the distribution of starting
points that I have explored include pseudo random
placement in the plane within a specified rectangular
region, linearly sequenced along a path (possibly even a
straight line), clustered within a cloud, or confined
within a narrow strip.
2.3. Variations and extensions
To better illustrate some of the issues discussed so far,
and to demonstrate the range of variations, additional
examples of drawings may be helpful. Figure 5 (top left)
shows an equally spaced set of points on a horizontal

Downloaded by [University of Windsor] at 06:02 19 November 2014

102

H. Dehlinger

Figure 5. Lines emanating from a linearly ordered sequence of points. Generative line drawing, pen and plotter. Copyright
H. Dehlinger.

line being used as starting points for the specimen lines


emerging from them. The other drawings in figure 5
make use of this same set of points. The four drawings in
figure 6 each use as the set of starting points a small
cloud of points. The lines in figure 6 are relatively long
and appear to fold into circular shapes due to their
segment rotation angle settings.
Although poly-lines are an inspirational and interesting artistic source for me, and they form the basis for
a lot of my experiments, I have used other line types
such as splines, parallel line pairs, short straight strokes
(similar to those used in figure 1), small circles, and
other small drawing element motifs to generate images.
Currently I am experimenting with algorithms that use
a line type (see figure 7 for an example) inspired by the
Inca Quipus.
Typical of the drawings I generate is that they tend
to use large numbers of lines in unusually dense
arrangements. The high density of these lines often
results in producing a grey scale effect. This positions
the genre of my respective drawings somewhere between
the medium of painting and the medium of line-art
drawing in the traditional sense.

2.4. Adding more complexity to line drawings


In order to generate drawings which include more
complex arrangements of lines, a rather large number of
starting points must be located in the plane for polylines to emanate from. The shapes of the starting point
configurations depend on the settings of the respective
initialization variables. Three examples of drawings that
take up this theme are shown in figure 8. The use of
failure prone equipment for the purpose of obtaining
new drawing effects may sound strange to someone with
a technical background, but it should not be very
surprising here. Failure introduces yet another form of
randomness different from the pseudo randomness
used by the algorithms that is even more difficult to
control. In figure 9 the detail of a drawing that was
executed using a faltering pen is shown. For the artist,
to invite such failures for the purpose of producing a
surprising new result is both acceptable and desirable.
The pen plotters I favour have now become almost
extinct as peripherals. They have been replaced by much
more convenient and faster printers. But plotter
technology has survived in the laser cutting industry.

103

Downloaded by [University of Windsor] at 06:02 19 November 2014

Generative line drawings

Figure 6. Line drawings with curling lines that all originate from the same set of starting points. Generative line drawing, pen and
plotter. Copyright H. Dehlinger.

The image in figure 10 shows a drawing burned into


cardboard using a laser.

3. Programming as a context for my drawings


Programming is very much an exercise in the art of
control because programs are expected to produce the
results they were programmed to produce. Programming
is logical and methodical. Clearly in todays world

computers and computer programs are useful, advantageous, and indispensable. They are a necessity. But there
is no necessity in art whatsoever. Artists operate in
a breathtaking space of epistemological freedom.
Everything is open to design. Artists who turn to
programming know why they want to do so. The
computer as machine acts as a deliberately chosen
contextual barrier. To process and calculate something
step by step is its great potential. The unstructured, the
only partially understood, the vague, the non-process,

104

H. Dehlinger

Downloaded by [University of Windsor] at 06:02 19 November 2014

Figure 7.

Experimental line-shapes inspired by Quipus, 2006. Generative line drawing. Copyright H. Dehlinger.

has no place. Considerable effort is necessary to achieve


satisfying results using programs. In my opinion, this is
one of the reasons that artists for whom programming
is a major aspect of their art production develop a body
of work with such a strong identity.
Besides me, algorithmic artists such as Hebert [8] and
Verostko [11] also focus mainly on the line and almost
exclusively use plotters. A recurring feature found in
many of Hebe`rts drawings expresses his concept of a
one-line-drawing. In such drawings an image that is
perceived as visually very complex, consists of a single
line only. Verostko was the first artist to mount a brush
on his plotting device. This fascinating idea of redesigning the hardware of a piece of precision equipment
intended for producing engineering drawings in order
to pursue an artists idea is a demonstration of how an
artist is trying to expand the given contextual boundaries to suit his concepts. Verostko also uses colours on
plotter drawings in a very distinct fashion. From this we
see that the way each of us formalizes what a line is
differs greatly.
The programmed line always requires precise specifications, and the act of deciding what those specifications are becomes the act where the intent of the artist
is introduced. This is what makes artist-programmed
lines so unique. Programmed lines convey the artists
personal expression and forge a recognizable identity.
For a generative line drawing, the conceptual intent of
the artist must first be expressed in a program, and only
then can the results be generated and evaluated. The
potential, the restrictions, the strengths and the weaknesses of the artistic concept are then manifest in the
program. An entire progression of value judgments has
to be made during the programming effort, and the
decisions made all contribute to clarifying the conceptual ideas and giving the finished work its identity.
I first started programming using Algol in the mid
1960s. I used Fortran for many years. I programmed the
CDC 6400 in Compass, and then learned SNOBOL 4 in

the early 1970s. My love of SNOBOL 4 continues to


this day and I still use it on my Macintosh. I have
programmed in LISP and Basic and used scripting
languages on CAD systems. More recently I have even
tried Java. The pen-plotter is my favourite output
device. It is a noisy mechanical device and it relies on
a real pen. It is master of lines only. But the line, the
simple line, has inexhaustible potential on a plotter. To
explore it is an artistic adventure.

3.1. More on the nature of the programmed line


When drawing a line segment there is an axiomatic
condition which cannot be violated: a straight line must
have a starting point and an ending point. Any drawing
device, human hand, or drawing machine has to obey
this axiom. When drawing a series of lines by hand the
following procedure must be invoked for each and every
line segment:
1.
2.
3.
4.

Set pen to starting point.


Draw line.
Remove pen from ending point.
If not done go to 1.

But this procedure does not describe what is going to


take place between each pair of decisions about where
to start and where to end. The hand receives this
information seriatim, one starting point and ending
point after another, as a result of a complex process into
which years of training have been invested.
A pen-plotting device operates in an almost identical
way. For example, the code required to draw a line
segment in the graphics programming language HPGL
[20] is:
1. PU x, y (move to position x, y with pen up)
2. PD u, v (move to position u, v with pen down)
The many sequential ad hoc decisions necessary for
drawing by hand that are made for the hand by the brain

105

Generative line drawings

Downloaded by [University of Windsor] at 06:02 19 November 2014

(a)

(b)

(c)

Figure 8.

Examples of generative drawings. Copyright H. Dehlinger.

Downloaded by [University of Windsor] at 06:02 19 November 2014

106

H. Dehlinger

Figure 9.

Detail of drawing with faltering pen, 1993. Copyright H. Dehlinger.

are replaced for the plotter by a pre-planned collection


of coordinate pairs stored in a data file. Thus line
segment generation becomes nothing more than a
formatted list of coordinate pairs representing start
points and end points. Their design requires careful
attention precisely because this is the only information
available. Another issue to consider is the characteristics
the lines will assume. Despite the fact that lines are
basically very simple geometric elements, they can
assume an infinite variety of shapes and appearances.
The way a line continues once it has left its starting
point, what expressive quality it will assume, what its
calligraphic character will be, and how it will interact
with other lines, are all examples of questions which
a generative drawing program will have to address. To
summarize, the important difference between drawing
by hand and drawing by making use of a program is:
by hand all decisions necessary are generated anew with
each line, whereas with a program all the decisions
necessary for drawing many thousands of lines can be
generated and stored so that the actual drawing process
can be a separate step.

4. Art, generative art, and the algorists


The question, What is art? is beyond the scope of this
paper, but since this question impacts my work, not to
address this question at all is of no help either. In my
view, to decide if an object O qualifies as art requires
that a judgment x which can assume the value O is art
be made by an authority A at time t in order to elevate
O to this status. Symbolically, one might formulate this
by writing
x fA O,t,
where the dependent variable x can assume only one of
two values. That is,
x 2 fart, :artg:
In this model art is declared to be art by an authoritative
act of judgment, and such authoritative acts are final.
Among all authorities, a particular Ais judgment may
be more accepted/rejected or less accepted/rejected
depending on Ais stature or weight. The object

107

Downloaded by [University of Windsor] at 06:02 19 November 2014

Generative line drawings

Figure 10. Drawing burned into cardboard with a laser, 2004. Copyright H. Dehlinger.

making process wherein the artist is guided by an artistic


concept usually, but not always, does include the
expectation of producing art that meets this criterion.
The artists concept is transformed into a production
process to yield a result that is offered for judgment.
However, sometimes a different purpose other than art
may govern the object generation process. For example,
much of the early art of Africa was not produced with
the intent of being art, but with the intent of being used
for ritual purposes. It was only declared to be art much
later. But just because time has elapsed there is no
guarantee that a transition of this kind will occur. The
author of a piece evaluates it at time t1, while others
may evaluate it at time tk44t1.
Where is the art in the drawings under discussion
here? Does it derive from the market value? Does it
derive from the use of a computer? Does it derive from
the content or message? Does it derive from the absence
of meaning? Does it derive from the fact that it is

presented in the familiar art format of a drawing? The


answer to all those questions is: No. It is art because it is
intended to be art and because it is offered for judgment
on that basis. The drawings considered here belong to a
specific genre of art. The genre is called generative art
and it makes use of algorithms. In a contribution to the
7th Generative Art conference, Galanter [21] defines
generative art as follows:
Generative art refers to any art practice where the
artist uses a system, such as a set of natural
language rules, a computer program, a machine, or
other procedural invention, which is set into
motion with some degree of autonomy contributing to or resulting in a completed work of art.
A procedure is a set of instructions for action, where
at each step it is clear what the next action will be. An
algorithm is a procedure that terminates after a finite
number of steps. It is not uncommon that artists decide to

108

H. Dehlinger

Downloaded by [University of Windsor] at 06:02 19 November 2014

Figure 11. An algorithm for the perspective projection method by Albrecht Durer (left), and a significant improvement of Durers
algorithm by the author (right). Copyright H. Dehlinger.

subject themselves to self-imposed restrictions that are


very close to what is understood as a program for a
computer. Action statements such as draw with short,
violent strokes, or go to and fro along a contour are
examples of artist self-imposed programs for drawing.
The generative process, as defined above, is quite general
and, as pointed out by Galanter [21], is not new to fine art.
Hebert [8], who as noted above also uses algorithms
in his work, defines an algorist as
. . . anyone in fact who invents and uses consistently one or several procedures to compose,
construct, produce, create art or to solve art
problems.
Hebert is the inventor of the JPH-algorithm which
defines an algorist in terms of an algorithm [22]:
if (creation && object of art && algorithm && ones
own algorithm)
{(include * an algorist * )}
elseif (!creation || !object of art || !algorithm || !ones
own algorithm)
{(include * not an algorist * )}
Further information on algorists and algorithmic art can
be found at [11]. Here, I simplify Galanters definition of
generative art while at the same time making it more
precise in the sense of Hebert by proposing the following
substitute definition:
Generative Art is an art practice where the artist
follows a self-designed formal system of rules.
5. A historical precedent for algorithms in art
The struggle by the Renaissance artists to master the
problem of perspective mapping from three-dimensional
space to two-dimensional space is an excellent historic

example of the use of algorithms in art. Albrecht Durer,


a German artist of the time, designed and published
several different mechanisms demonstrating algorithms
to solve this problem. His Draftsman of the Lute
woodcut depicts one of them [22]. In addition to his
woodcut Durer provides the following algorithm [23]
(freely translated by the author):
An assistant stretches a thread fixed to the wall
that runs through the picture plane, represented by
a wooden frame, behind which the master is
sitting, to a point on the Amphora. The master
mounts two threads to the frame so they touch the
thread the assistant is keeping taut. Then the
assistant loosens the thread he holds, closes
the door on the frame and the master transfers
the intersection point of the two threads he has
mounted on the frame to the door, which is opened
again by the assistant. Additional points are
determined in the same manner.
The left column in figure 11 shows the steps of
Durers algorithm, while my significantly improved
algorithm appears in the right column. My digital
woodcut in the manner of Durer that is shown in
figure 12 shows me posing in the role of the master. The
image uses elements from The Draftsmen of the
Amphora and the Draftsmen of the Lute mixed with
my own drawings. It seems unlikely that Durer ever used
the procedure he suggests in his woodcut Draftsman of
the Lute. He may, as Hockney [24] suggests, have
known about other methods to map three dimensions
into two dimensions. Currently, fast and efficient
algorithms to construct perspective views are a standard
component of CAD systems. Unfortunately, unlike
during the time of the Renaissance, today artists are
seldom able to contribute mathematical advances to
computational geometry.

Downloaded by [University of Windsor] at 06:02 19 November 2014

Generative line drawings

109

Figure 12. Digital woodcut showing the authors improved algorithm in the manner of Durer. Copyright H. Dehlinger.

6. The hand drawing and machine drawing universes


The famous novel The Library of Babel by Jose Luis
Borges [25] begins, The library (which some call the
universe) . . .. In the novel, Borges imagines and
describes a fantastic library, containing the totality of
all conceivable texts. Using this as a metaphor, we
imagine in a similar fashion both the Universe of Hand
Drawings and the Universe of Machine Generated
Drawings. These two universes of drawings contain the
totality of all conceivable drawings made by hand, and
the totality of all conceivable drawings made by
machines respectively. Each one of these universes is
well-defined and infinite. In such a universe, one should
expect to find hastily produced sketches, carefully
finished compositions, simple and complex drawings,
tasteless works, masterpieces, poetic works, etc. There
will also be pieces in each universe that mimic pieces
from the other universe. However, to be essentially
hand-like or essentially machine-like can be treated as
different qualities or distinguishing features. Of course
we can also envision subcategories in each of the two
universes. In the Universe of Machine Generated
Drawings, one interesting subcategory the one to
which the drawings discussed here belong is the
subcategory of algorithmically generated line drawings.
Figure 13 visualizes the two universes.
One can attempt to list some of the characteristics
which drawings in each universe will presumably exhibit.

For example, for the hand-like drawings one might


choose to focus on such features as
. no stroke will be identical to the preceding one;
. only with considerable effort will identical strokes
be possible;
. even when following a well-defined program
when making strokes, numerous small deviations
will occur and they will wind up being important for
the result;
. the position of the pen relative to the paper, the speed
of pulling, pushing, rubbing, the different pressure
conditions, and the motion of the will produce a range
of line attributes that will be difficult to formalize in
algorithmic, geometrical terms;
. the hand engaged in drawing lines does so under
control of the eye, which allows for direct feedback,
whose spectrum ranges from rational control of each
line to a completely vague perception of a wholeness
of doing;
. evidence of hand fatigue.
But for drawings generated by a machine one might
choose to focus on such features as
. the machine works untiringly, accurately, and fast
with continuous regularity;
. any irregularities occurring are due to failure conditions such as tearing of the paper, lack of ink
absorbency by the paper, dried-up pen, etc.;

Downloaded by [University of Windsor] at 06:02 19 November 2014

110

H. Dehlinger

Figure 13. The Universe of Hand Drawings and the Universe of Machine Generated Drawings, 1997. Copyright H. Dehlinger.

. pen pressure, pen speed, pen acceleration, and pen


position are constant;
. an algorithm is directing the pen and is unaware of the
result;
. only when the last line has been drawn can the result
be evaluated;
. the machine cannot examine or evaluate the result.
Concerning my algorithmically generated plotter
drawings, from an artistic point of view and from this
perspective, I offer the following statements regarding
my intentions.
. All lines must clearly reflect their machine-nature.
. It must be impossible, or at least difficult, for the
drawing to be done by hand.
. The lines must be drawn by a physical pen. (I view a
line drawn by a pen as being superior to a machine
printed line.)
. The plotting device, a device that was not
originally designed with the intentions of artists in
mind, should be pushed to its limits, and beyond, to
achieve an extraordinary density of lines within a
drawing.
. Failures of equipment, paper, pens, programs that
would never be tolerated in technical applications, are

viewed as valuable sources of potentially interesting


results.
. Each piece should stand on its own merits; how it was
made should be viewed as uninteresting.

7. Conclusions
In this paper I have provided a descriptive view of my
personal approach to generative line drawings. I have
described a kind of road map from which a
substantial body of work has been produced over the
last 25 years. I have not tried to compare different
mathematical concepts used in algorithmic art or to
critically evaluate the use of algorithms in art.
Although questions such as: Are some algorithms
better than others? and How do artists choose
algorithms? touch upon highly interesting theoretical
issues in generative art, they too have not been
addressed. Instead I have discussed and demonstrated
the interdisciplinary thinking and processes that
guide my algorithmic approach to art. I have given
a detailed view of the mathematical processes,
parameters and control mechanisms. I see my work
as embedded within the framework of the universes

Generative line drawings

that I have described. By this means I have come to


clearly see how humble my contribution is in the light
of the vastness of possibilities yet to be explored.
Programming for the purpose of creating art is a
pleasurable pastime. It boosts ones creativity; it
sharpens the senses and the mind; it offers one the
opportunity to experience success, defeat and frustration; it demands ones entire attention, but it also allows
for playful exploration; it is an exercise in mental
discipline; it teaches one to appreciate the surprising
consequences of randomness; and it offers hours of
Weltvergessenheit.

Downloaded by [University of Windsor] at 06:02 19 November 2014

Acknowledgements
I would like to thank Gary Greenfield for his valuable
suggestions for improving early drafts of this manuscript
and the anonymous referees for their many helpful
suggestions in revising this paper.

References
[1] http://www.emohr.com/ww4_out-htm (accessed 30 April
2007).
[2] http://www.atariarchives.org/artist/sec11.php (accessed
30 April 2007).
[3] http://www.dam.org/nake/ (accessed 30 April 2007).
[4] http://www.dam.org/nees/index.htm (accessed 30 April
2007).
[5] http://www.mgwilson.com/pltdraw.htm (accessed 30 April
2007).
[6] http://www.crca.ucsd.edu/hcohen/ (accessed 30 April
2007).
[7] http://www.csuri.com/ (accessed 30 April 2007).
[8] http://hebert.kitp.ucsb.edu/11s.html and: http://dam.org/
hebert/ (accessed 30 December, 2006).
[9] Krawczyk, R., 1999, Spirolaterals, complexity from
simplicity. In: N. Friedman and J. Barallo (Eds)
International Society of Art, Mathematics and
Architecture 1999 (San Sebastian, Spain: The University
of the Basque Country), pp. 293300.
[10] Krawczyk,
R.,
2001,
Curving
spirolaterals.
In: R. Sarhangi and S. Jablan (Eds) Bridges
Mathematical Connections in Art, Music, and Science
2001 (Winfield, KS: Central Plain Book Manufacturing),
pp. 2936.
[11] Verostko, R., 1990, Epigenetic painting: software as
genotype. Leonardo, 23(1), 1723. Available online at:
http://www.verostko.com/epigenetic.html (accessed 30
April 2007).

111
[12] Reichardt, J. (Ed.), 1968, Cybernetic Serendipity: Studio
International 168 (London: Institute of Contemporary
Art).
[13] Danilowitz, D., 2001, The Prints of Josef Albers: A
Catalogue Raisonne 19151976 (New York, NY: Hudson
Hills Press).
[14] Taylor, R.P., Micolich, A.P. and Jonas, D., 1999,
Fractal analysis of Pollocks drip paintings. Nature, 399,
422.
[15] Baume, N., 2001, Sol Lewitt: Incomplete Open Cubes
(Cambridge, MA: MIT Press).
[16] Dehlinger, H., 2005, Programming generative drawings
with small elements. In: C. Farsi (Ed.), Art and Math X
International Conference Proceedings, University of
Colorado, Boulder, 25 June, pp. 5155 (Boulder, CO:
University of Colorado).
[17] Foley, J.D., van Dam, A., Feiner, S.F. and Hughes, J.F.,
1990, Computer Graphics, Principles and Practice, 2nd edn
(Reading, MA: Addison-Wesley).
[18] Dehlinger, H., 2004, Generative art: fuzzy polygon
clipping in program generated line oriented drawings.
In: Gunther R. Raidl et al. (Eds), Applications of
Evolutionary Computing, Proceedings, EvoWorkshops
2004, Coimbra, Portugal, April 57, pp. 419426 (Berlin,
Heidelberg: Springer).
[19] Dehlinger,
H.,
1999,
Minimal
generative
principles for large scale drawings: an experimental
approach
and
its
Results.
In:
C.
Soddu
(Ed.), Proceedings of 2nd International Conference
Generative
Art,
Milan,
December
(Milan, Italy: Generative Design Lab of Politecnico di
Milano).
[20] HP 7475A Graphics Plotter, Interfacing and Programming
Manual, 1983 (San Diego, CA: Hewlett-Packard
Company).
[21] Galanter, P., 2003, What is Generative art?
Complexity theory as a context for art theory. In: C.
Soddu (Ed.) Proceedings of the 6th International
Conference on Generative Art, Milan, December, pp.
216235 (Milan, Italy: Generative Design Lab of
Politecnico di Milano).
[22] Rote, G., http://page.mi.fu-berlin.de/rote/Lere/2004-SS/
Computergrafik/Perspektive/D%FCrer-LauteBuch.jpg
(accessed 30 December, 2006). See also: http://de.wikipe
dia.org/wiki/Bild:358durer.jpg (accessed 30 December,
2006).
[23] Durer, A., 1525, Underweysung der messung mit dem
zirckel un richtscheyt in Linien ebenen und gantzen
corporen, durch Albrecht Durer zusammen gezogen und zu
nutz allen kunstliebhabenden mit zugehorigen figuren in
truck gebracht im jar MDXXV (2000, Berlin: Digitale
Bibliothek).
[24] Hockney, D., 2001, Secret Knowledge: Rediscovering the
Lost Techniques of the Old Masters (London, UK:
Thames & Hudson).
[25] Borges, J.L., http://jubal.westnet.com/hyperdiscordia/
library_of_babel.html (accessed 30 December 2006).

You might also like