In his thesis, Schmid explores new methods for tools that would give artists greater flexibility over the visual style of their computer animations. He studies stroke based rendering and develops two ways brush stroke rendering, specifically the brush stamping technique commonly used in image editors like Photoshop, can be applied to 3D digital renderings. His first method takes time and space into account to produce high quality images while his second takes hardware speed and capability into account to achieve high interactive rendering performance. These two methods both solve the problem of the depth order and occurrence order of paint strokes. His thesis develops not only only how to generate the 3D canvas that would hold the models, but also how to apply standard rigging tools and apply traditional 3D rendering processes (surface shaders, ray tracing etc).
Schmid defines a brush model as strokes represented by a geometric curve that would in essence act as the centerline of the stroke. From this skeleton curve, a procedural shader or ribbon texture recreates the appearance of a brush stroke. For the 2D platform, the representation of a brush stroke is brush stamping, in which a single brush texture is repeatedly blended along a curve.
The challenges of this technique in 3D, however, is difficult to represent as brush stamping does not capture how the shape and length of paint strokes change over time. Brush stamping also leads to aliasing problems (separated brush stamps causing lack of flow) that may lead to difficulties in rendering the 3D brush stroke. Combatting these issues, Schmid derives an algorithm that results in successful 3D representation results. When rendering brush stroke textures, transparency also plays a large role.
Moving on from brush strokes is canvas texturing. Canvas texture may refer to the texture of paper and enhances elements, such as brush stamps, placed on the canvas. Canvas texture is fixed with respect to the origin. Because “splats”, which refers to the textured quads rendered on the screen, are not fixed, moving brush strokes in front of the textured canvas create a “shower door” effect, in which brush strokes seem to be seen through a distortion layer, like a water droplets on a shower door. Canvas texture enhances the stylistic element of rendered brush strokes.
Schmid builds a 3D painting software system called OverCoat. By using a 3D canvas and his brush stroke techniques, paint strokes are embedded in space. Artists an create modeled proxy objects that define the overall layout of the scene. Geometric detailing is not required as strokes are embedded and the layout will not be rendered. OverCoat is based off mathematical optimization algorithms that I will not delve into as they are fairly complex and difficult to understand as well as explain. In addition to brush stroke rendering and canvas texturing, Schmid also includes various other tools that aid with animation and motion blurring. His software is demonstrated with rendered beautifully rendered 3D images that still retain stylistic brush strokes.
Schmid hopes to continue to build on his work. While OverCoat successfully creates 3D paintings, it also requires much manual work done by the artist in lighting and texture information. In the future, he wants to develop his software so that stylized 3D paintings can exhibit and support view-dependent shape changes that are reliant on dynamic lighting.
Schmid, Johannes. Methods of Artistic Stylization in 3D Animation. Thesis. ETH Zurich, 2012. Web.