32
Chapter 5
Conclusion
We present a technique for multi-material classification in the context of segmenta-
tion of density maps. Our classification method provides a smoother boundary than
found in typical segmentation. Furthermore, it enables sub-voxel flexibility in shape-
representation. This alleviates artifacts and jaggedness that can occur in visualization
of segmented volumes.
We have also developed an intuitive segmentation interface that uses a graph-cut
approach to combine user-knowledge and simplicity in segmentation. We have shown
that our tool can be effectively used to rapidly create segmentation.
For our future work, we will experiment with using higher level interpolants such as
В-splines for classification. This will give us greater flexibility and accuracy in defining
the boundary between materials. Additionally, we will examine segmentation in the
context of large cryo-electron tomography, which current techniques cannot address
due to size and fidelity of the volume.