Visualization of Knitwear

by Eduard Gröller, René Rau, Wolfgang Straßer

project duration: 1995-1996

This is a short description of a project that was done during a teaching and research visit of the first author at WSI-GRIS (University of Tübingen), Germany, dealing with the photorealistic rendering of virtual textile fabrics.


In Computer Graphics a boundary representation of objects is often used to generate realistic synthetic images. For certain objects it is more suitable to use a volume representation. In this project we investigate the visualization of knitting patterns where the details of the yarn were modeled by voxel data. The visualization and the modeling of textiles has already been investigated in depth in the computer graphics literature. However most of the publications consider woven fabrics. In the case of knitwear the topological specification of the knitting pattern allows a subdivision into basic elements. The thread course and the microstructure of the yarn are then approximated by voxel data. The periodic structure of knitting patterns allows a compact representation and an efficient rendering. The generated images demonstrate the possible realistic simulations.

Knitting yarns are made of twisted subyarns which themselves consist of microscopic fibres. Single fibres are not visible individually but contribute to the visual impression of the yarn structure. Yarns are therefore modeled as volume data where density values correspond to fibre frequencies.

Knitting patterns are typically characterized by a simple repetitive structure. Two types of stitches (plain stitch and reverse stitch) are used to produce such knitting patterns. Thus for a realistic simulation the knitting pattern is subdivded into basic elements as given in the following image.

The course of a knitting thread is modelled within such basic elements as 1D curve. A 2D thread cross section (density map) is moved along this 1D curve to produce the volume data set which represents a single basic element. The twisted shape of yarns is taken into account by rotating the thread cross section around the 1D curve.

Rendering of simple knitting patterns is done by volume visualization of the basic elements. In a subsequent step the resulting images are combined to give an image of the entire knitting pattern. Close-up views show that the 3D microstructure of knitwear is generated much better with our approach than 2D texture mapping techniques, an alternative approach, would produce.

The results achieved compare favorably (at least in our opinion) with scanned versions of real knitting patterns.

Here are some further computer-generated images:


Further information on the topic is available as

Additional information and extensions:

Links to other sites with research on knitwear, textiles:
Institute of Computer Graphics / Visualization and Animation Group / Research

This page is maintained by Eduard Gröller. It was last updated on November 20, 1998.
If you have any comments, please send a message to edi@cg.tuwien.ac.at.