Tracking curvilinear structures using active contours and application to microtubule videos

M. A. El Saban and B. S. Manjunath
Electrical and Computer Engineering Department, University of California, Santa Barbara
{msaban, manj} [at]


We present a novel active contour-based tracking algorithm for highly deformable and fast moving curvilinear structures. We propose to use line features as external forces for the active contour for the detection phase in the first frame. For the tracking part, we propose a novel approach of detecting motion of the endpoints and then deformation of the other points using the same active contour model used for detection. Estimating the motion at endpoints only has two advantages. First, motion estimation at endpoints is more robust at endpoints and corners, specially in noisy situations. Other points of the curvilinear structure may suffer from the aperture effect. Second, this will allow large deformation in the body of the structure from frame to frame. As an example application, we have tested our algorithm for biomolecular videos in tracking microtubules which exhibit high deformations in noisy video sequences. Experimental results show both quantitatively and qualitatively the good performance of the tracking algorithm.
[PDF] [BibTex]
M. A. El Saban and B. S. Manjunath,
VRL, ECE, UCSB, Jun. 2005.
Node ID: 438 , DB ID: 241 , Lab: VRL , Target: Technical Report