Stylized Vector Art from 3D Models with Region Support

Holger Winnemöller

Adobe Research

Elmar Eisemann

INRIA, Grenoble, France

John Hart

University of Illinois

David Salesin

Adobe Research

clipart_teaser1_000

This paper describes a process that converts polygonal meshed models of 3-D surfaces into the stylized vector-art 2-D layered polygonal depictions commonly found in popular clip-art libraries. In addition to contour extraction and processing common in previous work, this work also extracts, processes and stylizes regions. This focus on region relies on a novel combination of jagged, edge-based visual and shadow contours with smooth interpolated vertex normal shading contours. The process also maintains geometric precision by performing mesh operations in object space while using rasterization to accelerate topological classifications (e.g. visibility) of the contours and regions. The result is a stylized vector-art rendering system that enables consumers untrained in illustration to generate custom clip art.

Video & Results

newestProcess_000

This figure shows the process of converting a 3D mesh model into a stylized 2D vector image: First, we extract contours and then determine their visibility. We then construct a planar map and classify the regions enclosed by contours to yield a base vector depiction, which can then be stylized.

 

 

clipart_overview1

By separating content (3D model) from the style, we obtain many more combinations of possible visual results for a given model or style. Note that we can freely rotate the model, re-orientate the light-source, and change other style parameters. At the click of a button, the new vector graphic is computed within a matter of seconds.

Project Publications

Stylized Vector Art from 3D Models with Region Support

Eisemann, E., Winnemöller, H., Hart, J., Salesin, D. (Feb. 1, 2008)
Computer Graphics Forum (EGSR'08) , 27(4), 1199-1207