Image design by principles and random sampling

Jean Michel Morel and Luis Alvarez (JM&LA)

Un beau désordre est un effet de l'art

Nicolas Boileau (L’Art poétique, 1872)

What is abstract digital art?

Most digital images and videos created by humans are made by computer graphics methods. Because of its many applications to the entertainment and advertisement industry, computer graphics research is focused on modelling and imitating natural objects and environments. However, digital research on image structure can go far beyond the simple imitation of natural scenes. Humans feel the need to also explore all ways of creating new shapes and images, regardless of their plausibility in the real world. This is from early prehistory the goal of decorative art, and from the turn of the 20th century also the goal of abstract art. As observed in ancient pottery or in abstract art as well, giving up the imitation of nature leads to the search for general principles guiding the generation of (abstract, symbolic, decorative) shapes and images.

A quick examination of decorative arts and abstract design methods shows that most images and shapes are built by following a few simple generation rules. It is the choice of generating rules and of their combinations that defines an artistic or decorative style. This explains why modern artists sometimes used styles so similar that it was hardly possible to the layman to give a secure attribution. Ancient art, decorative art and abstract art have in common the search for abstract generating rules for shapes and images. In our age, such generating principles can be rephrased as algorithms creating digital images.

Painting without brushes: algorithms, rendering principles, and random generators

The creation act is shifted from imitating nature to the more abstract trend of inventing algorithms creating and combining shapes and images. Such algorithms can be automatized provided some of their choices are made randomly. In that way, the creation process no longer requires drawing or painting skills; the artist’s first activity is to program algorithms creating and combining shapes and colors on the one hand. Then, disposing of an unlimited number of digital images created effortless by a program, the artist’s main second activity is to choose by aesthetic criteria the appealing examples that were randomly generated, and to adjust a style to perfect effects that were at first observed by chance.

This is how we work. The complexity of our generated images derives from the systematic use of image generation principles borrowed from abstract art and decorative art such as occlusion, transparency, exclusion, inclusion, tessellation, symmetries, nonlinear deformations, etc... We use an extended version of our public online software http://www.ipol.im/aat. This software can work in fully automatic mode, where shapes, color palettes, and image generating principles are chosen randomly. Each image generation principle is steered by automatic random choices that come into play to decide of the colors of shapes, of their position, size, deformation and orientation in the frame, and of the interaction between randomly selected shapes. We chose to print our abstract generated images on paper and on silk scarves. Each model represents a style. To illustrate this fact, we have systematically made a series of each style, where each example is derived from the first by different random choices for the color palette and the shape disposition. The programs of each style are simple. They can be modified to perfect them, and they can be combined. Each new style generates virtually infinitely many different examples. This final generation is done almost effortless, but not quite so: a lengthy selection process takes place until a “right” realization is accepted by the artist. Each final composition is generally picked out of hundreds from the same style. We gave figurative names to some of the abstract compositions, like flowers or broken ice, based on remote visual reminiscences of natural scenes.

• 2019/06 DEMO release (implemented at IPOL (Image Processing Online Journal) DEMO site)
• 2018/07 DEMO release (implemented at IPOL (Image Processing Online Journal) DEMO site)
• Basic Online DEMO (old) (implemented at IPOL (Image Processing Online Journal) DEMO site)
• Online DEMO using background silhouettes (old) (implemented at IPOL (Image Processing Online Journal) DEMO site)
• Image Albums with Experiments
  • 2019 Design of silk scarves
  • 2019 Experiments width branched shapes II
  • 2019 Experiments width branched shapes I
  • 2014-2015 Experiments with symmetries and non-linear deformations
  • 2012-2014 Experiments with basic shapes and silhouettes
• Publications:
  • Luis Alvarez, Yann Gosseau, Jean-Michel Morel and Agustin Salgado. "Exploring the space of abstract textures by principles and random sampling", Journal of Mathematical Imaging and Vision 2015 (PDF)
  • Luis Alvarez, Nelson Monzon and Jean-Michel Morel. "Interactive Design of Random Aesthetic Abstract Textures by Composition Principles", accepted for publication in Leonardo (PDF)
• External related links:
  • Jean-Michel Morel Web Page