Public Programs
This page contains source programs recently coded by our laboratory.
Public Programs
Please read README for the details of each program. You can freely use the programs, but when you publish or present your work in which our programs are used, we are happy if you inform us so and publicly acknowledge so.
Ellipse fitting
Compact Algorithm for Strictly ML Ellipse Fitting
Given 2D points, this program computes the ellipse parameter using strict ML fitting.
References
K. Kanatani and Y. Sugaya, "Compact Algorithm for Strictly ML Ellipse Fitting", Proceedings of International Conference on Pattern Recognition (ICPR2008), Tampa, FL, U.S.A.
Ellipse fitting by hyper renormalization
Given 2D points, this program computes the ellipse parameter using the hyper renormalization.
References
K. Kanatani, A. Al-Sharadqah, N. Chernov, and Y. Sugaya, "Renormalization Returns: Hyper-renormalization and Its Applications", Proceedings of the 12th European Conference on Computer Vision (ECCV2012), Florence, Italy, 2012.
High accuracy ellipse-specific fitting
Given 2D points, this program computes the ellipse parameter using the combination of the hyper renormalization and RANSAC.
References
T. Masuzaki, K. Kanatani and Y. Sugaya, "High accuracy ellipse-specific fitting", Proceedings of the 6th Pacific-Rim Symposium on Image and Video Technology (PSIVT2013), Guanajuato, Mexico, 2013.
Fundamental matrix computation
High accuracy computation of rank-constrained fundamental matrix by efficient search (octave code)
Given 2D corresponding points over two images, this program computes the fundamental matrix using the Levenberg-Marquard method.
References
Y. Sugaya and K. Kanatani, "High accuracy computation of rank-constrained fundamental matrix by efficient search", Proceedings of the 10th Meeting on Image Recognition and Understanding (MIRU2007), July, 2007, Hiroshima, Japan, pp. 609-614.
Extended FNS for constrained parameter estimation (octave code)
Given 2D corresponding points over two images, this program computes the fundamental matrix using the extended-FNS method.
References
K. Kanatani and Y. Sugaya, "Extended FNS for constrained parameter estimation", Proceedings of the 10th Meeting on Image Recognition and Understanding (MIRU2007), July, 2007, Hiroshima, Japan, pp. 219-226.
Tracking and 3D recontruction
Given 2D trajectories of feature points in the image frame, this program reconstructs the 3-D shape of the scene by the projective reconstruction and computes the projection matirices for all the frames.
References
H. Ackermann and K. Kanatani, Fast projective reconstruction: Toward ultimate efficiency IPSJ Transactions on Computer Vision and Image Media, Vol. 49, No. SIG 6 (CVIM 20) (2008-3), pp. 68-78. (PDF)
H. Ackermann and K. Kanatani, Robust and efficient 3-D reconstruction by self-calibration Proceedings of the IAPR Conference on Machine Vision Applications (MVA 2007), May 2007, Tokyo, Japan, pp. 178-181. (PDF)
H. Ackermann and K. Kanatani, Iterative low complexity factorization for projective reconstruction, Proceedings of the 2nd Workshop on Robot Vision (RobVis08), February 2008, Auckland, New Zealand, pp. 153-164. (PDF)
Optimal triangulation
Triangulation from Two Views(C code)
Given 2D corresponding points over two images and their 3 x 4 projection matrics, this program optimally corrects the 2D corresponding points such that they strictly satisfy the epipolar equation.
References
K. Kanatani, Y. Sugaya, and H. Niitsuma, Triangulation from Two Views Revisited: Hartley-Sturm vs. Optimal Correction, 19th British Machine Vision Conference (BMVC2008), pp. 173--182.
Motion segmentation
Improved Multistage Learning for Multibody Motion Segmentation
Given 2D trajectories of feature points moving in the image frame, this program classifies them into independently moving motions, using the multi-stage unsupervised learning method initialized the GPCA. This algorithm first does initial segmentation by the GPCA. It progressively optimized by EM algoritm. The identity of each point (e.g, 0 for the background and 1 for a moving object) is returned to the standard output.
References
Multi-stage optimization for multi-body motion segmentation
Given 2D trajectories of feature points moving in the image frame, this program classifies them into independently moving motions, using the multi-stage unsupervised learning method. This algorithm does segmentation by unsupervised learning of degenerate motions followed by unsupervised learning of general motions. The identity of each point (e.g, 0 for the background and 1 for a moving object) is returned to the standard output.
Simulated and real image data (25.2MB)
References
Yasuyuki Sugaya and Kenichi Kanatani, Multi-stage optimization for multi-body motion segmentation, IEICE Transactions on Information and Systems, Vol. E87-D, No. 7, July 2004, pp. 1935-1942.
Yasuyuki Sugaya and Kenichi Kanatani, Geometric structure of degeneracy for multi-body motion segmentation, D. Comaniciu et al. (Eds.), Statistical Methods in Video Processing, Lecture Notes in Computer Science, No. 3247, Springer-Verlag, Berlin, December 2004, pp. 13-25.
Outlier removal for motion tracking by subspace separation
Given 2D trajectories of feature points moving in the image frame, this program detects false trajectories by robustly fitting an appropriate subspace to them and find those that have large residuals. The identity of each point (0 for outliers and 1 for inliers) is indicated in the standard output.
Real image data (51.6MB)
References
Y. Sugaya and K. Kanatani, Outlier removal for motion tracking by subspace separation, IEICE Transactions on Information and Systems, Vol. E86-D, No. 6 (2003-6), pp. 1095-1102.