Thesis Defense - Carlson

Surface Wrapping, a new approach to semi-automatic three-dimensional segmentation of volumetric data, can be used to rapidly and accurately extract object boundaries from volumes in which noise or poor imaging quality would otherwise make this process difficult. Current semi-automatic techniques often fail to perform effectively when the region of interest is both complex and poorly imaged, either requiring extensive editing of the resulting boundary or forcing the user to resort to manual segmentation. Surface Wrapping attempts to address these problems by combining a simple but robust interface for defining an arbitrarily complex approximate bounding surface with an interactive mesh deformation procedure, providing the best features of both manual and semi-automatic segmentation methods with few of the drawbacks of either.

This dissertation describes the design and implementation of Surface Wrapping, and evaluates the effectiveness of the approach using examples from both seismic and medical data sets. The technique is shown to be viable across a wide range of applications, with the potential to become a powerful addition to the existing array of 3D segmentation tools.