Abstract. Colour transfer algorithms aim to apply a colour palette, mood or style from one image to another, operating either in a threedimensional colour space, or splitting the problem into three simpler one-dimensional problems. The latter class of algorithms simply treats each of the three dimensions independently, whether justified or not. Although they rarely introduce spatial artefacts, the quality of the results depends on how the problem was split into three sub-problems, i.e. which colour space was chosen. Generally, the assumption is made that a decorrelated colour space would perform best, as decorrelation makes the three colour channels semi-independent (decorrelation is a weaker property than independence). However, such spaces are only decorrelated for well-chosen image ensembles. For individual images, this property may not hold. In this work, the connection between the natural statistics of colour images and the ability of existing colour transfer algorithms to produce plausible results is investigated. This work aims to provide a better understanding of the performance of different colour spaces in the context of colour transfer.

The statistics of natural images have attracted the attention of researchers in a variety of fields and have been used as a means to better understand the human visual system and its processes. A number of algorithms in computer graphics, vision and image processing take advantage of such statistical findings to create visually more plausible results. With this report we aim to review the state of the art in image statistics and discuss existing and potential applications within computer graphics and related areas. 1.