Nonlinear Color Triads for Approximation, Learning and Direct Manipulation of Color Distributions


We extend the notion of gradients and formulate nonlinear color triads that can represent a variety of naturally occurring color distributions that currently have no standard interactive representation. We derive a method to fit this this compact parametric representation to existing images and show its power for tasks such as image editing and compression. Further, we show that our formulation of the color triads can be included as a module in standard deep learning architectures, facilitating further research.

ACM Transactions on Graphics (SIGGRAPH) 2020
PDF BibTeX Project