Issue: Volume: 24 Issue: 8 (August 2001)

Painting with Feeling


In order to improve the art of digital painting, more attention needs to be focused on the craft of painting, according to researchers at the University of North Carolina at Chapel Hill. The art of painting refers to the aesthetic aspects of of the final product, while the craft of painting deals with the study of materials, including paint medium, tools, and methods.

To an artist, the art and craft of painting are closely related. Unfortunately, that connection has not been transferred to the computer graphics realm, according to UNC's William Baxter and Vincent Scheib, who, under the advisement of professors MingLin and Dinesh Manocha, have designed a novel interactive painting system to serve as a conduit to both art and craft. Rather than focus primarily on the rendered appearance of the final product, as is the case with most of today's digital paint tools, the new system, called dAb, pays due attention to re-creating the sight, touch, action, and feeling of the artistic process itself.

To incorporate such considerations, dAb uses a physically based, deformable brush model integrated with a haptic interface that is programmed to provide the force and tactile cues conceptually equivalent to real-world painting.

"The haptic stylus serves as a physical metaphor for the virtual paint brush," says Lin. "It takes in the position and orientation of the brush and displays the contact force between the brush and the canvas to the user." The brush bristles are modeled with a spring-mass particle system skeleton and a subdivision surface, enabling the creation of a large selection of brush types and a wide variety of strokes and painterly effects.

Developing the virtual 3D paintbrush required both a geometric representation and a physics-based model for its dynamic behavior. The design was constrained by the requirements of an interactive haptic painting system. "The real challenge was developing a simulation method that was stable enough and still computationally fast enough to maintain interactivity," says Baxter.
Digital art takes on a traditional appearance when generated with virtual paintbrushes that simulate the deformation of the bristles and the way paint is deposited on the canvas. UNC student James Ball painted this portrait with dAb.

"There are three simulations going on: the deformation of the brush, the spreading of the paint, and the haptics, which have to be fast enough so the artist doesn't perceive a lag," says Scheib.

For the brush simulation, the particle system reproduces the basic motion and behavior of a brush head, deforming as expected when colliding with the canvas. The deformable mesh skinned around the skeleton represents the actual shape of the head. Subdivision surfaces are particularly useful in this regard because of their ability to represent arbitrary topology and vertices easily.

By adjusting key parameters of the generalized 3D brush model, the researchers are able to create different types and shapes of brushes and mimic their physical behavior. For example, one type of brush is modeled as a single spine composed of a chain of particles. With this structure, the researchers can re-create a soft-style brush, such as those used in Japanese calligraphy, by employing a dynamics model that makes the brush appear more solid by eliminating oscillations. Stiffer brushes, such as those used in oil and acrylic painting, require a more complicated skeletal structure consisting of five mass particles connected with springs to form a pyramid. The four particles that form the base are rigidly fixed to the brush handle and are directly driven by the user's input. The fifth particle serves as the point of the brush.
The dAb interface consists of a brush, a palette with paint, and a digital canvas. Users can mix the available colors to create complex color gradients as well as entirely new colors.

For the haptic expression and processing, dAb uses a SensAble Phantom and a dual-processor Pentium III PC with an Nvidia GeForce2 card. One processor is dedicated to the force computation and the other is used to simulate brush dynamics as well as paint transfer and blending. With the haptic system, the user paints directly onto a virtual canvas displayed on the screen. Using the space bar as a toggle, the user can bring up the virtual palette for paint mixing and brush cleaning or put the palette aside to paint on the canvas.

Critical to the effectiveness of the 3D system is its ability to provide sufficiently good force feedback to emulate the sensation of applying brush strokes to a canvas, says Baxter. "We align the virtual paintbrush with the six degree-of-freedom stylus and position it so the point of three degree-of-freedom force output coincides with the point where the head meets the handle on the virtual brush." The system is driven by a basic force model that computes compressive effects and frictional forces.

The final component of the dAb system is a multi-layered paint model that supports such features as bi-directional paint transfer (brush to canvas to brush), blending, drying, glazing, and loading various portions of the brush head with different pigments. While these effects can be achieved with other advanced painting programs, notes Scheib, "with our paint model, the user only has to manipulate the virtual brushes [in the same manner] as real ones in order to automatically generate the intended paint effects."

In addition, the 3D virtual brush enables the user to mix and create a nearly unlimited number of new colors from the available colors, as well as to create complex color gradients on the digital palette by mixing the constituent colors less thoroughly.

Encouraged by the positive user feedback so far, the researchers are considering ways to extend the technology. "Just about every component of the system could be improved: better brush modeling, better paint modeling, better force feedback," says Baxter. "The challenge is to balance such improvements with the amount of computation required. It's quite easy to kill the interactivity of the system with simulation algorithms that are too computationally intense. Among the potential enhancements are more efficient integration and simulation methods to improve the accuracy of the brush deformations and the simulation of a greater range of haptic phenomena, from the feel of paint textures to the variation in sensation when using different types of brush fibers. Another natural step, according to the researchers, is the application of the technology to painting 3D geometric models.

dAb is one of many ongoing projects in the UNC GAMMA research group funded by the Army Research Office, Intel, the National Science Foundation, and the Office of Naval Research. More information can be found on the project Web site at