Henrik Lieng    Christian Richardt    Neil A. Dodgson

Computer Laboratory, University of Cambridge, United Kingdom

Computational Aesthetics 2012

Left: Colours distributed using Matlab’s random-number generator. Middle: Result of our algorithm.
Right: An image similar to one of Damien Hirst’s spot paintings. Notice that our algorithm does not produce any distinctive patterns.


Apparently-random distributions of colours in a discrete setting have been used by many artists and craftsmen in the past century. Manual colourisation is a tedious and difficult process. Automatic colourisation, on the other hand, tends not to not look ‘random’ to a human, as randomly-generated clusters and patterns stimulate human perception and break the appearance of randomness. We propose an algorithm that minimises these apparent patterns, making the distribution of colours look as if they have been distributed randomly by a human. We show that our approach is superior to current solutions, especially for small numbers of colours. Our algorithm is easily extendible to non-regular patterns in any coordinate system.



This is the authors’ version of the work. The definitive version is available at http://diglib.eg.org/.


    author    = {Henrik Lieng and Christian Richardt and Neil A. Dodgson},
    title     = {Random discrete colour sampling},
    booktitle = {Proceedings of Computational Aesthetics},
    year      = {2012},
    month     = jun,
    location  = {Annecy, France},
    doi       = {10.2312/COMPAESTH/COMPAESTH12/081-087},
    url       = {http://richardt.name/coloursampling},