Fitting Multiple Connected Ellipses to an Image Silhouette Hierarchically

In this paper, we seek to fit a model, specified in terms of connected ellipses, to an image silhouette. Some algorithms that have attempted this problem are sensitive to initial guesses and also may converge to a wrong solution when they attempt to minimize the objective function for the entire ellipse structure in one step. We present an algorithm that overcomes these issues. Our first step is to temporarily ignore the connections, and refine the initial guess using unconstrained Expectation-Maximization (EM) for mixture Gaussian densities. Then the ellipses are reconnected linearly. Lastly, we apply the Levenberg-Marquardt algorithm to fine-tune the ellipse shapes to best align with the contour. The fitting is achieved in a hierarchical manner based upon the joints of the model. Experiments show that our algorithm can robustly fit a complex ellipse structure to a corresponding shape for several applications.

[1]  P. Fua,et al.  Tracking Of Hand ’ s Posture And Gesture , 2004 .

[2]  W. Eric L. Grimson,et al.  Adaptive background mixture models for real-time tracking , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).

[3]  Richard Y. D. Xu,et al.  Multiple curvature based approach to human upper body parts detection with connected ellipse model fine-tuning , 2009, 2009 16th IEEE International Conference on Image Processing (ICIP).

[4]  Erkki Oja,et al.  Randomized hough transform (rht) : Basic mech-anisms, algorithms, and computational complexities , 1993 .

[5]  Bogdan Kwolek,et al.  Stereovision-Based Head Tracking Using Color and Ellipse Fitting in a Particle Filter , 2004, ECCV.

[6]  Andrew Blake,et al.  Articulated body motion capture by annealed particle filtering , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).

[7]  Andrew W. Fitzgibbon,et al.  Direct Least Square Fitting of Ellipses , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[8]  Paul L. Rosin.  A note on the least squares fitting of ellipses , 1993, Pattern Recognit. Lett..

[9]  Richard Szeliski,et al.  Matching 3-D anatomical surfaces with non-rigid deformations using octree-splines , 1993, Proceedings of IEEE Workshop on Biomedical Image Analysis.

[10]  Nikolaos Grammalidis,et al.  Head detection and tracking by 2-D and 3-D ellipsoid fitting , 2000, Proceedings Computer Graphics International 2000.

[11]  Jeff A. Bilmes,et al.  A gentle tutorial of the em algorithm and its application to parameter estimation for Gaussian mixture and hidden Markov models , 1998 .

[12]  Richard Y. D. Xu,et al.  An iterative approach for fitting multiple connected ellipse structure to silhouette , 2010, Pattern Recognit. Lett..

[13]  Pascal Fua,et al.  Tracking articulated bodies using Generalized Expectation Maximization , 2008, 2008 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops.

[14]  J. M. Hans du Buf,et al.  Contour profiling by dynamic ellipse fitting , 2000, Proceedings 15th International Conference on Pattern Recognition. ICPR-2000.

[15]  Rolf P. Würtz,et al.  Organic Computing Methods for Face Recognition (Methoden des Organic Computing zur Gesichtserkennung) , 2005, it Inf. Technol..