Development of an Integrated Computer Art System
 

Published in New Trends in Computer Graphics, Springer Verlag, pp. 643 - 652, 1988, ISBN 0-387-19328-6

3,300 words



 
mike king >> writings >> Development of an Integrated Computer Art System
mike king| postsecular | jnani
writings | graphics | cv
 


   

Abstract

An 'Integrated Computer Art System' is described which offers the computer artist, illustrator and designer a much wider range of 2D manipulations than in conventional paint systems. The system allows non-programmers to explore a range of techniques based on computer geometries and algorithms, including certain types of fractals, in an interactive fashion, while still giving full control down to the pixel level. The system is under development, but already shows potential for generating new computer imagery not belonging to the now 'conventional' world of 3D photo-realism.

Introduction

Computer graphics has long been associated with 3D scene simulation, with increasing compute- power devoted to the techniques of photo-realism. For the artist, illustrator and graphic designer these methods have limited use, and they have found the 2D paint system, with all its limitations, to be more suited to their budgets, needs and abilities. Paint systems have a well established nature by now, and do not receive the intensive research effort that photo- realism is still receiving. In my proposed 'Integrated Computer Art System' (ICAS) I am trying to redress the balance by extending the paint system concept with an increased functionality that addresses the needs of the non-programming computer artist and designer. ICAS, in brief, provides a mixture of paint and object-oriented techniques with additional features including pattern-making, tesselations and interactive fractals. The range of facilities have been selected by a careful consideration of what the computer can offer to the artist and designer working in two dimensions, based on a range of old and 'new' geometries. The system can also be used to create three dimensional imagery, but not based on the usual computing techniques of photo- realism. These are too restricting to the artist, for whom a personal feel for depth, light, shade and perspective is what gives their work a unique character.

1) The concepts behind ICAS

In a paper entitled Towards an Integrated Computer Art System (King 1987) I outlined the basis for ICAS, building on my PhD research (King 1986), while this paper presents the results of implementing some of those proposals. The PhD thesis concluded that the computer provides for a range of new media in the visual arts, and that their user interfaces are an essential part of the media. In the thesis I developed a taxonomy of computer art media, and some new terms which describe concepts that I feel are important in this area, and I will briefly summarise these here.

The first distinction that I have made is between interactive and scripted systems. In interactive systems the user is not required to program the system, and most of the image-making involves a high level of hand-eye feedback, as in a paint system. Scripted systems, in contrast, imply the creation of a program script, usually text-based, and its subsequent execution for the realisation of the imagery. Feedback is thus through a write - execute - rewrite cycle.

The creation of imagery in either type of system is by synthesis from primitives, except where image processing or equivalent techniques are used, in which case a degree of image analysis is also needed. The concept of synthesis from primitives is a very important one in this context, and is what makes art and design applications of computers unique. Most computing applications revolve around information retrieval and analysis, with the reduction of a large data base into a few concise conclusions a frequent goal. Alvey Ray Smith (1984) recognised this when he coined the term 'data base amplification' for the techniques behind a range of computer generated imagery. The creative and synthetic aspects of computer art and design systems has not received a large amount of attention however, and the concepts that I am introducing here should help to make a more detailed discussion of these applications possible.

In the context of synthesis from primitives I make a second important distinction: between arbitrary and algorithmic synthesis from primitives, and indeed much of what follows depends on this distinction. Arbitrary synthesis from primitives involves a sequence of operations on the medium that derive from the artistic whim or intuition of the user - the machine has no 'understanding' of the sequence or control over it. Algorithmic synthesis involves a sequence of operations governed by a set of rules that are communicated to the machine and encapsulated in a machine-executable algorithm. As a simple example the incorporation of set or coloured pixels following a free-hand curve in a paint system is arbitrary, while the incorporation of pixels in a (Bresenham) straight line is algorithmic. One of the challenges of any art medium is the balance between algorithmic and arbitrary techniques, and I believe it to be especially so with computer art. In looking at algorithms used by computer artists and computer graphics researchers since the sixties to generate imagery, I have realised that these can all be considered as embodiments of different types of geometry. Looking more closely at these algorithms or geometries, it seemed that they could conveniently be classified as 'classical' or 'recursive' geometries, and I have listed a 'computer artist's geometrical toolkit' under these two headings:

The Computer Artist's Geometrical Toolkit

classical GEOMETRIES
· geometries of parallel lines, triangles, rectangles and polygons
· the conic sections: circle, ellipse, parabola and hyperbola
· nets, bands and tesselations
· non-recursive functions
· Lissajou's figures, cardioids and cycloids
· parametric curves

RECURSIVE GEOMETRIES

· iterative functions (recurrence relations)
· random numbers
· recursive patterns
· fractals and graftals
· particle systems
· growth models
· linear and array grammars
· Markov chains

2) Proposed structure of ICAS and progress to date

The proposed structure of ICAS, outlined in King (1987), was a two-level system, with an object-oriented or drafting level from which one could 'descend' into a paint level. The drafting level was to provide the usual object-oriented manipulations, but greatly extended in terms of pattern-making, tesselations and fractals (recursively substituted shapes). The paint mode allows for pixel-editing of the output from the drafting mode, raster-based cut and paste, and more gestural and expressive free-hand work.

In the object-oriented or drafting mode the intention was to implement a wide range of the geometries listed above in the toolkit'. However, because one can never anticipate all the possible needs of the user, and because some geometries are difficult or clumsy to control interactively, it is intended that ICAS should include a scripted component, which would manipulate the image through the creation of small 'scripts' or programs.

Development effort so far has concentrated on the integration of a drafting medium and a paint system. The drafting aspect has allowed for a comprehensive range of manipulations not easily provided for in paint systems, and also the implementation of the sophisticated pattern-making facilities, and some interactive fractals. The system is written in C and runs on IBM PC compatibles driving an Io Research Ltd 8 bit per pixel frame store. The data structure for the drafting mode is a doubly-linked list of nodes, each containing a field identifying the primitive (line, poly-line, rectangle, filled rectangle, circle etc.), and fields containing the variables of instantiation (position, size and colour). The sequence of the linked list determines the order of drawing, so that for the user to change the layering the sequence has to be changed by changing the pointers in the list. Superimposed on the list is a tree-structure reflecting the hierarchy that the user imposes on the drawing by creating groupings. The tree-structure is implemented using grouping nodes that are pointed to by the elements in the group, and that point back in turn to the elements. The group nodes also contain information on the extents of the group. An individual group is manipulated by recursively descending its structure until primitives are reached.



3) Work-table philosophy

Normally in a paint system all the user's input is stored as a modification to the array of pixels in a frame buffer, and recovery from the latest modification is only possible if multiple picture planes are available (these are expensive and give only limited recovery). Scaling up or down results in a sampling effect and picture degradation, and hence the full screen tends to be in use for the image in a paint system. Whatever the resolution, the user is unlikely to want to give up portions of the screen 'real-estate' to menus, palettes, and icons representing various 'tools'. These tend to be stored off-screen and blitted onto the screen when required. This means that spare frame-buffer capacity is required for interactive systems, and is of course expensive. One of the tricky parts of designing a paint system on a limited budget, is that often there is little spare memory.

With an object-oriented or drafting system the problems of scaling up and down are reduced because of the underlying (vector) description of the picture. In the design of ICAS I have therefore used only some 80% of the screen for the image in drafting mode, while using the entire screen for paint mode. The advantage of working in the drafting mode is that all the tools are at hand, as illustrated in Fig. 1.

Fig. 1 Screen layout of the ICAS 'work-table'

I call this a 'work-table' approach, because it more closely parallels a traditional artist's work table, on which one would have to hand brushes, paints, geometrical instruments and so on. The paint system approach of blitting up the palette and menus for every single change of colour, brush and operation is a bit like keeping all these items in a drawer under your table, with the added complication that opening the drawer obscures part of the art-work! This is not to be taken as a criticism of windows in general; it is just an awkward way to work in this particular application. Coming back to the work-table philosophy, the work area can in fact be used to store many motifs, scraps of design and so on. The ability to reduce a fragment down and place it in a corner for later use adds greatly to the power of the system. While the work table is useful in the construction of an image, it is a nuisance in that the shapes and colours in the menus and palettes are distracting from the overall composition and effect of the created image. Artists get round this by pinning a work in progress to a blank wall where one can more easily judge the composition. In ICAS this need is taken into account by a facility that removes the surrounding menus and palettes and re-draws the image to full scale - as it would be seen on exit to paint mode. A press of a key then returns the user to the work table with the reduced image for further manipulation. One would typically go in and out of the work table a number of times in the creation of a finished product, in order to judge its progress.

4) Design and motif creation

The basic techniques while in drafting mode are similar to those found in Macdraw - a drafting system for the Apple Macintosh. The common techniques of rubber-banding, dragging, and echoing the selection of an item by highlighting are provided, along with a method for capturing items for manipulation as a group. Because the palette is always visible, smooth shaded polygons can be easily created by selecting a new colour between vertices. Flood-fills are provided, and can be individually deleted by traversing a stored list of them which highlights them in turn. Figures 2 to 9 show the sequence of creation of some images, and also illustrates the range of imagery that the system is capable of. Some of the images contain lines that have been roughened by a function provided within the fractal menu (described later).

Fig. 2
First part of a design created using mirror, rotate and copy

Fig. 3
Design reduced and rotated and more elements added

Fig. 4
A series of roughened lines are created

Fig. 5
The previous designs are combined using cut and paste, and free-hand elements added in paint mode

Fig. 6
First stage of a design using shaded polygons

Fig. 7
The polygons are grouped as a whole, scaled, mirrored and rotated, and other elements added

Fig. 8
The final image cut and pasted in paint mode

Fig. 9
Mostly drafting techniques are used in this abstract composition


5) Pattern-making

I regard pattern-making as one of the classical computer geometries, and an obvious area for exploitation by computers in art and design. At present various arrangements of motifs can be displayed, based loosely on a classification in MacGregor and Watt (1984). The user designs a motif, usually to a large a scale as possible, and then scales it down for repeating. Because of the data-structure underlying the motif the scaling down can be safely performed, secure in the knowledge that it will scale up perfectly. The work-table philosophy comes into play again with pattern-making, because only the currently selected grouping is patterned. The pattern covers the entire screen, the work-table at first disappearing On the work-table can be any number of motifs and scraps of design, and of course these all re-appear on return to drafting mode. Figure 10 shows the ICAS work-table with some motifs and the following figures show some sample patterns.

Fig. 10
ICAS work-table with some motifs for patterning

Fig. 11
A simple pattern

Fig. 12
A combination of patterns


The system has been used by myself and a few students so far, and it is surprising how fast one can experiment with different motifs and groupings. Important elements of pattern-making are rapidly made apparent: the difference of the final effect with the same motif but with different groupings, and for any given motif and grouping the arising of emergent forms. These are shapes created by the gaps between the elements of the motif, and also the longer thrusts of pattern, which range from harmonious to disturbing in their overall effect.

Tesellations are not yet implemented, but will again be done at the drafting level. A tessellation differs from the patterns created through the 17 network groups in a few essential ways. The key to a tessellation is its outline, not just its orientation. One can think of a tessellation as consisting of a plane covered with identical blank tiles, whose outlines fill the plane without gaps, and whose outlines mark out the image. A form that tessellates may vary from a rectangular tile to the intricate figurative shapes of Escher, and may or my not be decorated with motifs. If the form for tessellation is decorated with a motif there is no requirement that the motif should interlock - it is just the outline that is important. Figure 13 shows the basis on which ICAS will provide interactive tesselations. As with patterns the tessellation will be generated full-screen for exit to paint mode, or, on return to the work-table, the complete tessellation can be incorporated into the data-structure for subsequent manipulation.

Fig. 13 Interactive tesselations

6) Interactive fractals

All the previous facilities so far described are based on what I have called the classical geometries. I have also introduced some recursive geometries into the system with the provision of some simple fractals. These are based on the recursive substitution of a collection of line segments (called the generator) into another collection of line segments (initiator). At its simplest this is provided by a function called roughen which uses a simple built-in generator consisting of 3 line segments. More interesting fractals can be created by the use of user-defined generators however. As with pattern-making and tesselations the system 'goes out' to full screen to draw the selected fractalised initiator. The resulting image depends on the sense of both initiator and generator - i.e. whether they are drawn clockwise or anticlockwise, leading in the case of the Von Koch snowflake to the normal or anti- snowflake (see Mandelbrot, 1982). The generated fractal image can be 'incorporated' on return to the work-table, for manipulation in ways already described. Figure 14 shows some examples.

Fig. 14
Fractal trees cut and paste with recursive rectangles

As with most algorithmic synthesis from primitives, once the degree of 'data base amplification' goes over a certain (fairly low) level, the results (first time) are unpredictable, though deterministic. The nature of the results can be anticipated with experience, and it is exactly this experience that ICAS can so rapidly provide to the artist and designer, both with classical and recursive techniques.


7) Exit to Paint

On 'exit to paint' the palettes and menus are removed and the image drawn full-scale, or, in the case of patterns, tesselations and fractals, the current selection only is 'patterned out' to full size. The image is now manipulated at a paint level, with the normal low level of editing. Colour and look-up table manipulation is provided, and indeed the paint component of ICAS is full- functioned enough to be sufficient for the needs of many artists. I have found with some of my students that both classical and recursive geometries hold little interest for them at first; they find the expressive capabilities of TV resolution 8-bit painting quite sufficient, and sufficiently different from conventional media to be exciting. However, the point of ICAS is to extend the paint system to include techniques that are unique to the computer, either because the technique allows increases of orders of magnitude in speed of execution (object-oriented manipulations, pattern-making and tesselations), or because the technique is in practice impossible to realise by hand (for example fractals).

8) The next step: constraint-driven geometries

The concept of using constraints to describe geometrical problems to the computer goes back to Sketchpad, through Smalltalk (Byte 1981) and ThingLab (Borning 1979) to Greg Nelson's Juno system (Nelson 1985). I plan to make a constraint-solver the scripted component of ICAS. I conceive the scripted component of ICAS to be closely integrated with the rest of the system and to look very little like a conventional high-level programming language. The ideas put forward by Greg Nelson and his forerunners will be the starting point, with the interesting challenge of developing constraint-driven recursive geometries. Imagery from chaotic functions (one of the recursive geometries in my classification) could be generated by scripts of a few lines.
Conclusion

ICAS is something like half-way towards fulfilling the original concept and has already shown interesting results, not just in term of the range of imagery produced with the system, but also in terms of user interface design. I believe that the development of such systems, taking place outside the research thrust into photo-realism in computer graphics, will have an impact on the visual language of a new generation of artists and designers. By extending the paint system with increased control over both old and new geometries, the non-programming artist has direct access to a new world of imagery.

References

Borning 1979: Borning, Alan, "Thinglab - a Constraint-Oriented Simulation Laboratory" SSL- 79-3, Xerox PARC, Palo Alto CA July 1979.
King 1986: King, M.R., "Computer Media in the Visual Arts and their User Interfaces" PhD thesis, Royal College of Art, London 1986.
King 1987: King, M.R., "Towards an Integrated Computer Art System" in Earnshaw, R.A. and Lansdown, R.J., (eds.) "State of the Art in Computer Art and Animation" conference proceedings, Springer Verlag, 1987.
MacGregor and Watt 1984: Macgregor, J. and Watt, A., "The Art of Microcomputer Graphics" Addison-Wesley Publishing Co., 1984.
Mandelbrot 1982: Mandelbrot, B.B., "Fractals: Form, Chance and Dimension" W.H.Freeman and Company.
Nelson 1985: Nelson, G., "Juno, a constraint-based graphics system" Computer Graphics, Volume 19, Number 3, 1985, pp. 235-243.
Smith 1984: Smith, A.L., "Plants, Fractals and Formal Languages" SIGGRAPH 84, pp. 1-10.






 
mike king >> writings >> Development of an Integrated Computer Art System
mike king| postsecular | jnani
writings | graphics | cv