We present a computational model to predict perceptual coarseness from an image. The model was based on the hypothesis that a model for predicting perceptual coarseness should be motivated by human visual system. We found that the amplitude of the Fourier transform of an image captures information of coarseness. Furthermore, an analysis of Fourier amplitudes in terms of the human contrast-sensitivity function (CSF) leads to a metric that can predict perceptual coarseness. Model performance was proved by comparing the perceptual coarseness that was predicted by the model from images of metallic paint panels and the psychophysical data which was obtained by a visual assessment using the physical panels.
S. Kitaguchi, M.R. Luo, E.J.J. Kirchner, G.J. van den Kieboom, "Computational Model for Perceptual Coarseness Prediction" in Proc. IS&T CGIV 2006 3rd European Conf. on Colour in Graphics, Imaging, and Vision, 2006, pp 278 - 282, https://doi.org/10.2352/CGIV.2006.3.1.art00056