A Fast Active Contour Tracking Method Based on Gaussian Mixture Model

Conference paper
Part of the Lecture Notes in Electrical Engineering book series (LNEE, volume 246)


This paper proposed a Gaussian mixture model based gradient level set method (GMM-GLS) for moving target contour tracking in video sequences to handle automatic initialization and background variation. In contrast with conventional level set models, adaptive GMM background subtraction is applied to get the rough location of moving target as foreground in current frame. And more accurate mask image according to the rough location of foreground with dilatation operation indicates the initialization contour of level set evolution. Then, the gradient level set model can evolve the curve quickly and ensure more accurate convergence to the target contour in tracking procedure. Based on this accurate mask, the GMM-GLS method can greatly reduce the uncertain iteration time in curve convergence and optimize the initialization of GLS eliminating the interferential background. Experimental results on many real-world video sequences validate that our approach greatly improves the performance of object contour tracking.


Geometric active contours Level set Gaussian mixture model 


  1. 1.
    Kass M, Witkin A, Terzopoulos D (1987) Snakes: active contour models. Int J Comput Vis 1:321–331. doi: 10.1007/bf00133570 CrossRefGoogle Scholar
  2. 2.
    Xu C, Prince J (1998) Snakes, shapes, and gradient vector flow. IEEE Trans Image Process 7:359–369. doi: 10.1109/83.661186 CrossRefMATHMathSciNetGoogle Scholar
  3. 3.
    Caselles V, Catte F, Coll T, Dibos F (1993) A geometric model for active contours in image processing. Numer Math 66:1–31. doi: 10.1007/BF01385685 CrossRefMATHMathSciNetGoogle Scholar
  4. 4.
    Caselles V, Kimmel R, Sapiro G (1997) Geodesic active contours. Int J Comput Vis 22:61–79. doi: 10.1023/A:1007979827043 CrossRefMATHGoogle Scholar
  5. 5.
    Osher S, Sethian JA (1988) Fronts propagating with curvature dependent speed: algorithms based on Hamilton-Jacobi formulation. J Comput Phys 79:12–49. doi: 10.1016/0021-9991(88)90002-2 CrossRefMATHMathSciNetGoogle Scholar
  6. 6.
    Weickert J, Romeny BMH, Viergever MA (1998) Efficient and reliable schemes for nonlinear diffusion filtering. IEEE Trans Image Process 7:398–410. doi: 10.1109/83.661190 CrossRefGoogle Scholar
  7. 7.
    Li C, Xu C, Gui C (2005) Level set evolution without re-initialization: a new variational formulation. Proc IEEE Conf Comput Vis Pattern Recogn 1:430–436. doi: 10.1109/CVPR.2005.213 Google Scholar
  8. 8.
    Mumford D, Shah J (1989) Optimal approximation by piecewise smooth functions and associated variational problems: commun. Comm Pure Appl Math 42:577–685. doi: 10.1002/cpa.3160420503 CrossRefMATHMathSciNetGoogle Scholar
  9. 9.
    Chan T, Vese L (2001) Active contours without edges. IEEE Trans Image Process 10:266–277. doi: 10.1109/83.902291 CrossRefMATHGoogle Scholar
  10. 10.
    Rathi Y, Vaswani N, Tannenbaum A, Yezzi A (2005) Particle filtering for geometric active contours with application to tracking moving and deforming objects. IEEE Comput Soc Conf Comput Vis Pattern Recogn 2:2–9. doi: 10.1109/CVPR.2005.271 Google Scholar
  11. 11.
    Sun X, Yao H, Zhang S (2011) A novel supervised level set method for non-rigid object tracking. IEEE Conf Digital Object Identifier Comput Vis Pattern Recogn 3393–3400. doi:  10.1109/CVPR.2011.5995656
  12. 12.
    Zivkovic Z (2004) Improved adaptive Gaussian mixture model for background subtraction. Proceedings of the 17th international conference on pattern recognition, vol 2, pp 23–26. doi: 10.1109/ICPR.2004.1333992Google Scholar
  13. 13.
    The School of Informatics of the University of Edinburgh (2004) CAVIAR Test Case Scenarios. Accessed 31 Apr 2013

Copyright information

© Springer International Publishing Switzerland 2014

Authors and Affiliations

  1. 1.School of Communication and Information EngineeringShanghai UniversityShanghaiChina
  2. 2.Shanghai Key Laboratory of Digital Media Processing and TransmissionsShanghai Jiao Tong UniversityShanghaiChina

Personalised recommendations