Yoshimichi OKADA*1), Takeshi KOISHI*2), Suguru USHIKI*2)
Toshiya NAKAGUCHI*1,2), Norimichi TSUMURA*1,2), Yoichi MIYAKE*1,2,3)
*1 Graduate School of Advanced Integration Science
*2 Graduate School of Science and Technology, Chiba University
*3 Research Center for Frontier Medical Engineering, Chiba University
Medcal Imaging Technology, Vol.22, No.1, pp.35-40 (2004)
Abstract
In this paper, we propose fast stereo matching algorithm for 3D reconstruction
of internal organs using a stereoscopic laparoscope. Stoyanov et al. proposed
a technique that recovers the 3D depth of internal organs from the images
of a stereoscopic laparoscope. In this conventional technique, the dense
stereo correspondence was solved based on the registration of the whole
image regions in order that the 3D depth of internal organs can be densely
recovered. However, the calculation cost of this technique was very high
since the registration of whole image regions requires multidimensional
optimization. In this paper, we propose algorithm using a local-region
based registration method considering the similarity of stereo images taken
by a stereoscopic laparoscope. The local-region based registration requires
only low dimensional optimization. Therefore, its calculation cost is significantly
low. We evaluated the calculation cost of the proposed algorithm using
a stereoscopic laparoscope and abdominal models of a human body. We also
evaluated the accuracy of the algorithm using a 3D laser scanner and abdominal
models. The results showed that the proposed algorithm can significantly
reduce the calculation cost with a little accuracy degradation of 3D depth.
Full paper(to appeare)
[PDF] ( in Japanese )