Evolutionary Search for the Artistic Rendering of Photographs

This chapter explores algorithms for the artistic stylisation (transformation) of photographs into digital artwork, complementing techniques discussed so far in this book that focus on image generation. Most artistic stylisation algorithms operate by placing atomic rendering primitives “strokes” on a virtual canvas, guided by automated artistic heuristics. In many cases the stroke placement process can be phrased as an optimisation problem, demanding guided exploration of a high dimensional and turbulent search space to produce aesthetically pleasing renderings. Evolutionary search algorithms can offer attractive solutions to such problems. This chapter begins with a brief review of artistic stylisation algorithms, in particular algorithms for producing painterly renderings from two-dimensional sources. It then discusses how genetic algorithms may be harnessed both to increase control over level of detail when painting (so improving aesthetics) and to enhance usability of parameterised painterly rendering algorithms.