Improving medial surfaces for reverse engineering

Common medial axis transform methods create two artifacts: shrinkage of the end of surfaces and small dent at junctions. They are due to the definition of medial axis transforms and they are bottlenecks in industrial applications. This paper presents two methods to improve the medial surfaces computed from CT images of thin-plate mechanical objects. We first introduce a method for extending the ends of medial surfaces by finding the intersection points of medial voxels on the isosurface. We also propose a method to remove dent structures at junction edges by minimizing error functions. These improvements, integrated into a conventional medial surface extraction method using sub-sampling, resolve several drawbacks in reverse engineering applications.