The aim of this paper is to describe a method for transforming an image's 3D colour histogram so as to accurately match a predetermined target state. The method proposed here consists of a colour indexing stage followed by the determination of a histogram transformation matrix on the basis of the Earth Mover's Distance histogram difference metric.1 In addition to describing this method, the paper also analyses the results of transforming an image's 3D colour histogram in terms of its effect on the image's spatial characteristics. Finally, examples of using the technique to transform an image's histogram to match that of another image are shown. The purpose of having developed this approach is to be able to perturb this image characteristic in an accurate, direct and controlled way as this can be of use in studies that aim to study the impact of image characteristics on various imaging contexts, like colour reproduction or database indexing and searching. Being able to modify the 3D colour histograms of an image then allows for the generation of image test sets in which images have this characteristic in arbitrary states.
Ján Morovic, Pei–Li Sun, "Transforming 3D Colour Histograms of Images" in Proc. IS&T CGIV 2002 First European Conf. on Colour in Graphics, Imaging, and Vision, 2002, pp 104 - 108, https://doi.org/10.2352/CGIV.2002.1.1.art00024