Colour Spaces for Colour Transfer

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.

Sliced and Radon Wasserstein Barycenters of Measures

“…We consider a color image represented as a vector X ∈ R N×3 of N pixels, so that X = (X k ) k=1,...,N where each pixel X k ∈ R 3 stores the value of a pixel indexed by k. In the following, we use the YCbCr color space because of its ability to decorrelate color channels, although other color spaces may be used (e.g., the CIE-Lab advocated in [28]). The color distribution of this image is a measure µ X defined in R 3 , and describes the color palette.…”

Water quality and recreational use of public waterways

“…That people benefit from access to natural and manmade water bodies is well documented (Reinhard and Pouli 2011;Kistemann 2011, 2013). Increasing the visibility of blue space in urban areas has been associated with lower psychological distress (Nutsford et al 2016) and there is evidence that exposure to blue spaces during physical activity shortens perceived exercise duration and increases willingness to repeat such exercise in the future (White et al 2015).…”

Colour normalisation of histopathological images

“…For better quality, colour space must be decorrelated which makes the three colour channels semi-independent (Reinhard and Pouli 2011). Reinhard and Pouli (2011) showed that the choice of colour space is important. For example, the RGB colour space produces poor colour transfer results as compared to Lab colour space.…”

“…For these images, the concept of swatches is proposed, in which the user has to select some swatches in the source and target images and has to apply the colour transfer method between the corresponding swatches (Reinhard et al 2001). The other colour spaces used for colour transfer methods are CIELab; Yxy; Yuv; Yu 0 v 0 ; XYZ; RGB and HSV (Reinhard and Pouli 2011). But these colour transfer methods do not produce good results for histopathological images as they contain a large number of different objects.…”