An interface for sketching three-dimensional (3D) wireframe models using a picture (two-dimensional image) as a guide is described in this paper. Suggestive techniques are used to provide hints to users about possible subsequent operations, inferred from the current context and user gestures.
Three-dimensional sketching  and predictive (suggestive) interfaces  have been described in earlier research. This paper augments those ideas by adding the use of an existing image as a basis for the 3D wireframe model to be created.
The system described in the paper uses curves and lines as the drawing elements. Gestures (done using a stylus device) are used for interaction instead of the traditional windows, icons, mouse, pointer (WIMP) interface.
Ongoing suggestions for the possible geometry to use are provided to the user in the same viewing space (an idea similar to that provided in some word processors and source code editors to help users complete a word). Suggestions are provided for possible curve closure, and for possible extrusion. Choices are also presented from a database of geometry that was previously used for a similar curve to the current one.
Image-guided 3D sketching, which is the main contribution of the paper, is achieved by allowing the user to import images onto the three orthogonal construction planes. Images are assumed to be line sketches. Image filtering techniques are used to snap the user-traced curve to the matching features in the underlying image. Operations to avoid undesirable snapping, and to specify points on the curve (called gluing and pinning in the paper) are also supported.
The work described is interesting, but in practice, the effort and the training required will prevent any bridging of the gap between the sketch artist (speed) and the 3D modeler (accuracy). On the other hand, educational software, games, and drawing packages can benefit immensely from the use of the techniques described in the paper.]]