Interactive algorithm for accurate segmentation of fiber-like objects

This paper presents a semi-automated algorithm for accurate and reproducible segmentation of fiber-like objects. Novel features are introduced that effectively characterize fiber-like objects. The method is based on the classic live-wire algorithm, but two important innovations are suggested for decreasing user introduced bias: optimization of manually edited seed points (automated set-up of seed points) and automated generation of seed points. We illustrate the performance of the method on computer generated synthetic images as well as on real biomedical images.