UC Davis : Computer Science : IDAV : Oliver Staadt : Research : Virtual Clay Modeling

Virtual Clay Sculpting

Abstract

An ongoing project is research into virtual clay sculpting. This project is attempting to find a volumetric approach to clay modeling that looks visually realistic, can handle non-trivial sized datasets, and can run at interactive rates. Our target frame-rate is to achieve at least 10 frames per second.

We are using a signed distance field to represent the clay, with only the interior of the clay having defined values. The advantage of using a signed distance field to represent the clay is that we have a defined gradient at every point within the clay volume, which is useful for many possible clay transport schemes. The surface of the clay is implicit in the zero-set of the signed distance field.

The primary challenge with this scheme is to reinitialize the signed distance field quickly when the user's actions have caused the surface of the clay to change. The majority of the work on this project has been related to finding a fast way to solve this. The final solution involves identifying only those cells whose values could have changed as a result of the user's actions, and then reinitializing only those cells. The result is that the speed of the algorithm is dependent on the number of cells affected by a given user action, rather than the size of the entire dataset. Our technique does add some additional error to the data, but the error is detectable, and is on the order of error generated by traditional reinitialization techniques. Most importantly, our technique makes our clay sculpting system possible.

I am still in the final stages of completing a 2D version of the clay sculpting system, and am working to eliminate instabilities in the data. Once this is completed, the way forward to a 3D version is relatively clear, and the system has been designed with the 3D case in mind. We expect, based on preliminary results, to be able to allow user edits on regions of at least 30x30x30.

Publications

A poster describing the fast reinitialization scheme developed as part of this work is being presented at Computer Graphics International 2005. A PDF abstract of that poster is available here.

Contact

Eric Klein elklein@ucdavis.edu
Oliver Staadt staadt@cs.ucdavis.edu