Three Ways of Interpreting a Line

One of the questions I am most interested in exploring with KIKAIMA X MICHI is how to translate an image into a drawing. At first, the system simply identified the contours of a photograph and turned them into lines the machine could reproduce on paper. The results were interesting, but all the lines had the same thickness regardless of what they represented, and I needed a way to test different brush pressures on the paper, which would produce lines of different thicknesses.

To explore this possibility I have developed three different methods that allow line thickness to change.

The first is based on the concentration of parallel lines. When interpreting an image, several marks can appear very close to one another and follow similar directions. Rather than treating them only as independent lines, the system can interpret their accumulation as an area needing greater graphic presence. It is a way of relating the density of the marks to their importance in the drawing.

The second method uses contrast. A photograph contains transitions between lighter and darker areas, some much more pronounced than others. I am interested in using that information so a line can vary in thickness along its path. A very clear separation between two surfaces could thus produce a thicker mark, while a softer transition could be represented by a finer line.

The third method, which I call structural persistence, starts from a different idea. If we progressively blur a photograph, small details begin to disappear, but some forms and contours remain recognizable. The system compares these different versions of the image to identify which lines remain and which are lost. Based on that information, it can give greater presence to contours that remain visible and resist the loss of information as the image degrades and its details fade.

What interests me about these three methods is that each interprets the importance of a line differently. A line can matter because it is accompanied by other nearby lines, because it separates two areas of strong contrast or because it belongs to a structure that remains recognizable when details disappear. And those criteria do not necessarily agree. A highly contrasted line may be a small detail, while a less pronounced one may belong to a much more recognizable form. There is no single answer as to which should have greater presence in the drawing.

The three methods can be used separately or together, allowing me to experiment with different interpretations of the same image. I do not expect the software to understand what it sees as a person would, but to develop criteria through which it can construct a graphic language I find interesting.

Then there is still the question of bringing those thicknesses onto paper. On screen I can represent a line with a particular width, but when I use a brush its shape, pressure, ink and the properties of the support come into play. The physical result does not necessarily correspond exactly to what the program shows.

I intend to continue exploring these methods and see what results they produce with different tools and materials. I am especially interested in finding out whether combining different ways of interpreting an image can bring me closer to a more personal form of drawing. Ultimately, I do not want MICHI to reproduce a photograph as accurately as possible, but to interpret it through decisions that form part of my own pictorial language.