Feature aware image warping and 3D-Collages November 7, 2006

In this talk I would like to present two recent projects that I have been involved in. The first project deals with feature- aware image warping that is useful for 2D and 3D texture mapping. We present a method for inhomogeneous image warping guided by a feature mask that preserves some regions of the image, such as foreground objects or other prominent parts. Our mapping may receive an arbitrary warping function as input, and produce its feature-aware version that preserves the shape of the image features as much as possible. In particular, our method allows global or local changes to the aspect ratio of the image without causing undesirable shearing to the features.
In the second part of the talk, I will discuss our ongoing work on non-realistic shape modeling. Inspired by the paintings of Arcimboldo (1527-1593) and modern surrealist art, we are developing a tool for computer-aided creation of expressive 3D collages, where a target shape is modeled by a collection of smaller elements, such that both the whole and the parts are simultaneously recognizable and establish a semantic relationship envisioned by the artist. The tedious task (for humans) of performing partial shape matching and placement of objects in 3D is performed by our algorithm, allowing the artist to concentrate on the creative side of the work.

Official Website: http://cg.cs.tu-berlin.de/~sorkine/