Adaptive Scale Mean-Shift Tracking with Gradient Histogram

  • Changqing Xie
  • Wenjing Kang
  • Gongliang LiuEmail author
Conference paper
Part of the Lecture Notes in Electrical Engineering book series (LNEE, volume 516)


The mean-shift (MS) tracking is fast, is easy to implement, and performs well in many conditions especially for object with rotation and deformation. But the existing MS-like algorithms always have inferior performance for two reasons: the loss of pixel’s neighborhood information and lack of template update and scale estimation. We present a new adaptive scale MS algorithm with gradient histogram to settle those problems. The gradient histogram is constructed by gradient features concatenated with color features which are quantized into the 16 × 16 × 16 × 16 bins. To deal with scale change, a scale robust algorithm is adopted which is called background ratio weighting (BRW) algorithm. In order to cope with appearance variation, when the Bhattacharyya coefficient is greater than a threshold the object template is updated and the threshold is set to avoid incorrect updates. The proposed tracker is compared with lots of tracking algorithms, and the experimental results show its effectiveness in both distance precision and overlap precision.


Object tracking Mean-shift Scale estimation Gradient 



This work was supported by the National Natural Science Foundation of China (Grant No. 61501139) and the Natural Scientific Research Innovation Foundation in Harbin Institute of Technology (HIT.NSRIF.2013136).


  1. 1.
    Fukunaga K, Hostetler L. The estimation of the gradient of a density function, with applications in pattern recognition. IEEE Trans Inf Theor. 1975;21(1):32–40.MathSciNetCrossRefGoogle Scholar
  2. 2.
    Comaniciu D, Ramesh V, Meer P. Real-time tracking of non-rigid objects using mean shift. In: Proceedings IEEE conference on computer vision and pattern recognition, 2000. IEEE; 2000, vol. 2. p. 142–9.Google Scholar
  3. 3.
    Comaniciu D, Ramesh V, Meer P. Kernel-based object tracking. IEEE Trans Pattern Anal Mach Intell. 2003;25(5):564–77.CrossRefGoogle Scholar
  4. 4.
    Collins RT. Mean-shift blob tracking through scale space. In: Proceedings IEEE computer society conference on computer vision and pattern recognition, 2003. IEEE; 2003, vol. 2. p. II-234.Google Scholar
  5. 5.
    Vojir T, Noskova J, Matas J. Robust scale-adaptive mean-shift for tracking. Pattern Recogn Lett. 2014;49:250–8.CrossRefGoogle Scholar
  6. 6.
    Canny J. A computational approach to edge detection. In: Readings in computer vision; 1987. p. 184–203.Google Scholar
  7. 7.
    Sobel I. An isotropic 3 × 3 image gradient operator. In: Machine vision for three-dimensional scenes; 1990. p. 376–9.Google Scholar
  8. 8.
    Wu Y, Lim J, Yang MH. Online object tracking: A benchmark. In: 2013 IEEE conference on computer vision and pattern recognition (CVPR). IEEE; 2013. p. 2411–8.Google Scholar
  9. 9.
    Grabner H, Grabner M, Bischof H. Real-time tracking via on-line boosting. Bmvc; 2006, vol. 1, no. 5. p. 6.Google Scholar
  10. 10.
    Oron S, Bar-Hillel A, Levi D, et al. Locally orderless tracking. Int J Comput Vision. 2015;111(2):213–28.MathSciNetCrossRefGoogle Scholar
  11. 11.
    Henriques JF, Caseiro R, Martins P, et al. Exploiting the circulant structure of tracking-by-detection with kernels. In: European conference on computer vision. Springer, Berlin, Heidelberg; 2012. p. 702–15.CrossRefGoogle Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2020

Authors and Affiliations

  1. 1.School of Information and Electrical EngineeringHarbin Institute of Technology (Weihai)WeihaiChina

Personalised recommendations