Study of Specific Location of Exhaustive Matching in Order to Improve the Optical Flow Estimation

  • Vanel Lazcano
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 738)


Optical flow is defined as pixel motion between two images. Hence, in order to estimate optical flow, an energy model is proposed. This model considers: a data term and a regularization term. Data term is an optical flow error estimation and regularization term imposes spatial smoothness. Most of traditional variational models use a linearized version of data term, which fails when the displacement of the object is larger than their own size. Last years the precision of optical flow method has been increased due to the use of additional information, which comes from correspondences computed between two images obtained by: SIFT, Deep-matching or exhaustive search. This paper presents an experimental study to evaluate strategies for locating exhaustive correspondences improving flow estimation. We considered different location for matching: random location, uniform location, maximum of the gradient and maximum error of the optical flow estimation. Best performance (minimum EPE and AAE error) was obtained by the Uniform Location which outperforms reported results in the literature.


Motion estimation Large displacement Variational model Gradient constancy constraint Color constancy constraint 


  1. 1.
    S. Baker, D. Scharstein, J. Lewis, S. Roth, M. Black, R. Szelinsky, A database and evaluation methodology for optical flow. Int. J. Comput. Vis. 92, 1–31 (2011)Google Scholar
  2. 2.
    T. Brox, C. Bregler, J. Malik, Large displacement optical flow, in IEEE conference on Computer Vision and Pattern Recognition (2009)Google Scholar
  3. 3.
    A. Bruhn, J. Weickert, C. Feddern, T. Kohlberger, C. Schnoerr, Real-time optical flow computation with variational methods, in International Conference on Computer Analysis of Images and Patterns 2003, The Netherlands (2003), pp. 222–229Google Scholar
  4. 4.
    B.K.P Horn, B.G. Schunck, Determining optical flow. Artif. Intell. 17, 185–204 (1981)Google Scholar
  5. 5.
    V. Lazcano, Some problems in depth enhanced video processing, Ph.D. thesis, 2016.
  6. 6.
    J. Sánchez, E. Meinhardt-Llopis, G. Facciolo, TV-L1 optical flow estimation. Image Process. Line 3, 137–150 (2013).
  7. 7.
    M. Smith, R. Hashemi, L. Sears, Classification of movies and television shows using motion, in 6th International Conference on Information Technology: New Generations, Las Vegas, Nevada (2009)Google Scholar
  8. 8.
    F. Steinbruecker, T. Pock, D. Cremers, Large displacement optical flow computation without warping, in IEEE International Conference on Computer Vision (ICCV), Kyoto (2009), pp. 185–203Google Scholar
  9. 9.
    M. Stoll, S. Volz, A. Bruhn, Adaptive integration of features matches into variational optical flow methods, in Asian Conference on Computer Vision - ACIP2012 (2012)Google Scholar
  10. 10.
    Y. Wang, K. Gurule, J. Wise, J. Zheng, Wavelet based region duplication forgery detection, in 12th International Conference on Information Technology: New Generations, Las Vegas, Nevada (2012)Google Scholar
  11. 11.
    P. Weinzaepfel, J. Revaud, Z. Harchaoui, C. Schmid, DeepFlow: large displacement optical flow with deep matching, in IEEE International Conference on Computer Vision (ICCV), Sydney (2013)Google Scholar
  12. 12.
    A. Wedel, T. Pock, C. Zach, H. Bischof, D. Cremers, An improved algorithm for TV-L1 optical flow, in Statistical and Geometrical Approaches to Visual Motion Analysis. Lecture Notes in Computer Science, vol. 5604 (Springer, Berlin, 2009)Google Scholar
  13. 13.
    L. Xu, J. Jia, Y. Matsushita, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), San Francisco, CA, 13–18 June 2010Google Scholar

Copyright information

© Springer International Publishing AG, part of Springer Nature 2018

Authors and Affiliations

  1. 1.Núcleo de Matemáticas, Física y Estadística, Facultad de CienciasUniversidad MayorSantiagoChile

Personalised recommendations