“Computer-Generated Watercolor” describes the ways in which computer science researchers managed to approximate the effects of watercolor painting. This was actually a really exciting paper to read, because it involves art and math that I actually understand (yay vector calculus!).
The researchers used a three-layer model to simulate the fluid dynamics of watercolor interacting with paper. The top layer is the shallow water layer, where the water and pigment flow above the surface of the paper. The middle layer is the pigment deposition layer, where the paper absorbs or desorbs the pigment. The bottom layer is the capillary layer, where water traverses the paper by capillary action.
The researchers also modeled several varieties of watercolor paper with different textures. Each of these has its own height and fluid capacity values.
Much of the paper consists of describing the math (spoiler alert: gradients and partial derivatives everywhere) and programming behind the manipulation of the water and pigment, and how to simulate the conditions that produce recognizable watercolor effects. It also discusses the visibility of various pigments on white or black paper, and how the researchers used the Kubelka-Mink model (which I am unfamiliar with) for optical compositing of the layers.
Here are some examples of real watercolor effects:
Here are the same examples produced with computer-generated watercolor: