Cellular Automata and Non-Photorealistic Rendering

Rachel Bryant


Supervised by Paul L Rosin; Moderated by Xianfang Sun

Cellular automata (CA) consist of a regular grid of cells, each of which can be in only one of a finite number of possible states. The state of a cell is determined by the previous states of a surrounding neighbourhood of cells and is updated synchronously in discrete time steps. The identical rule contained in each cell is essentially a finite state machine, usually specified in the form of a rule table with an entry for every possible neighbourhood configuration of states automata use simple rules which are able to generate complex patterns

Image based Non-photorealistic Rendering (NPR) processes images using image processing and computer vision techniques to re-render them in a more artistic style.

This project will combine both these techniques to perform NPR image stylisation for colour images using CA.

Since colour images will be used, then it is not possible to enumerate all possible patterns in a neighbourhood. Therefore, an alternative approach will be taken where a set of rules will be predefined in an algorithmic way (cf. totalistic CAs) The system will use training pairs of images (original image plus NPR version) to learn which set of rules to apply to produce results similar to the training example results.

Final Report (27/11/2020) [Zip Archive]

Publication Form