Vesselness Based Feature Extraction for Endoscopic Image Analysis
Digital Object Identifier (DOI)
Distinctive features are crucial to many tasks in computer assisted minimally invasive surgeries (MIS). Most existing methods are difficult to extract distinctive features in MIS images. For better analysis of MIS images, we resort to blood vessels that are abundant and distinctive on the tissue surfaces. Based on vascular branching points, we propose a new type of vascular feature, branching segment. Two novel methods, Vesselness Based Circle Test (VBCT) and Vesselness based Branching Segment Detection (VBSD) are proposed to detect branching points and branching segments respectively. In the experiments, the performance of VBCT and VBSD is evaluated with in vivo images and VBCT is compared with other state-of-the-art feature point detectors. The numerical results verify that branching points and branching segments are highly repeatable under different viewpoints. Moreover, the computational complexity of VBCT and VBSD is linear to the number of pixels. As supplements to other types of feature point detectors, VBCT and VBSD provide researchers new tools for endoscopic image analysis.
Was this content written or created while at USF?
Citation / Publisher Attribution
2014 IEEE 11th International Symposium on Biomedical Imaging (ISBI), Beijing, p. 1295-1298.
Scholar Commons Citation
Lin, Bingxiong; Sun, Yu; Sanchez, Jaime; and Qian, Xiaoning, "Vesselness Based Feature Extraction for Endoscopic Image Analysis" (2014). Computer Science and Engineering Faculty Publications. 89.