Simulating Chinese Brush Painting

Calligraphy by Ma Qingxiong, my brush painting teacher.

Project Drink Ink

News
For the full thesis in PDF click here.

I'm presenting a poster at SIGGRAPH 04. Download the poster here. Poster (PDF)

For the SIGGRAPH poster abstract click here.

Abstract

Since 1986, when Steve Strassmann of MIT pioneered the simulation of black ink painting in his paper "Hairy Brushes", researchers in the field of non-photorealistic rendering (NPR) have been developing more faithful systems for brush painting. Chinese brush painting presents a particularly interesting challenge due to the unique properties of the ink, paper, and brush. In the last 3 years advances in the simulation of ink diffusion in paper have brought realistic simulation of brush painting closer to reality. Ink diffusion alone, however, if not enough to produce aethetically pleasing digital paintings. The brush needs to be acurately modeled in order to capture the subtle details that produce complex and beautiful brush strokes. Strassmann's hairy brush model represented the brush as a linear array of fibers that deposit ink particles on the paper as the brush moves. Since the development of this basic model, 3D dynamic simulations have been used to create far more realistic brushes that deform as a result of pressure of friction. The results of recent brush modeling work is visually stunning as seen in the "Virtual Chinese Brush" project [http://www.cs.ust.hk/~cpegnel/VCB/]. This project suggests that the brush model is more important to the aethetic quality of the painting than the ink diffusion model. The "Virtual Chinese Brush", however, requires a special 6-DOF input hardware in order to feed the brush the required data. This makes it unsuitable for automatically rendering realistic brush strokes over curves derived from 3D models or input with a mouse. My thesis project will focus on creating a more realistic model of the brush and couple this with a system that interpolates pressure, angular orientation, and velocity from any path in a style that emulates traditional painting technique.

Development Screen Shots

		Finally a new screen capture (-:.
		Finally a new screen capture (-:. A lot has changed since the last one in terms of how euclid works. I'll update the abstract soon to reflect this. The double brush is an artifact from taking a couple of screenshots. It seems that my dynamic texture composited the screenshot with itself...oops.
		Using a mouse, some very poor calligraphy was created. The real input device should improve upon this greatly simply by allowing the movements to be natural.
		The painting begins. With painfully primitive input controls the brush can now paint.
		Test modeling of ink using a simple RGBA texture map that is formed from compositing the texture map of the previous time step with the bristles intersecting the paper at the current time step.
		The next step is to build a 3D, 6-DoF input device. Giovanni Motta is assisting me in the construction of a system that will use orthogonally placed webcams to track LEDs embedded in an actual brush.
	A Novel Method of Modeling a Soft Brush: Consider the total energy of each bristle as the sum of the energies from friction, bending, and twisting. Now constrain each bristle to be above the paper. Since a bristle is defined as a polyline, it is simple enough that we can intuit the mimimum energy state given the constraint. Now find an explicit geometric representation for this inuitive mimimal energy state. Implications of this Approach: Because I'm deriving a geometric model for the bristles, it is possible to model the brush as a collection of thousands of bristles and still calculate state transitions in realtime. Physical based approaches cannot achieve this level of complexity and perform in realtime.