B.J. Kadlec and H.M. Tufo (USA)
medial surface, structure tensor, distance transform
Medial surface extraction is a powerful tool for compact shape description, feature tracking, surface generation, and other image processing and visualization applications. Current techniques are ineffective for processing noisy realistic data, many are difficult to implement, and others only focus on 2D data. In addition, no previous approaches utilize the distance transform for segmentation of the medial surface. In this paper, a robust and simple 3D algorithm for extracting and segmenting the medial surface from arbitrary three-dimensional objects is presented. Critical points in the level sets of an object’s distance transform are used to locate the medial surface. The structure tensor is used for both the extraction and to constrain a hierarchical segmentation technique for partitioning the medial surface into meaningful components. The proposed extraction method is computationally efficient, simple to implement, robust on noisy and complex topologies, and performs surface segmentation with little additional cost. The technique is tested on arbitrary 3D objects to demonstrate correctness and applied to extracting and segmenting faults from noisy seismic data to show robustness and real-world applicability.
Important Links:
Go Back
