Extended Lucas-Kanade Tracking

Shaul Oron, Aharon Bar-Hillel, Shai Avidan


The Lucas-Kanade (LK) method is a classic tracking algorithm exploiting target structural constraints thorough template matching. Extended Lucas Kanade or ELK casts the original LK algorithm as a maximum likelihood optimization and then extends it by considering pixel object / background likelihoods in the optimization. Template matching and pixel-based object / background segregation are tied together by a unified Bayesian framework. In this framework two log-likelihood terms related to pixel object / background affiliation are introduced in addition to the standard LK template matching term. Tracking is performed using an EM algorithm, in which the E-step corresponds to pixel object/background inference, and the M-step to parameter optimization. The final algorithm, implemented using a classifier for object / background modeling and equipped with simple template update and occlusion handling logic, is evaluated on two challenging data-sets containing 50 sequences each. The first is a recently published benchmark where ELK ranks 3rd among 30 tracking methods evaluated. On the second data-set of vehicles undergoing severe view point changes ELK ranks in 1st place outperforming state-of-the-art methods.


  • Extended Lucas-Kanade Tracking PDF BIB
    Shaul Oron, Aharon Bar-Hillel, Shai Avidan
    European Conference on Computer Vision, 2014

Spotlight Video (ECCV 2014)

NOTE: Video has sound !

Tracking Examples

Tracking examples for two sequences taken from the tracking benchmark of Wu et al. CVPR 2013.
Videos show bounding box predicted by the tracker the current template and the object/background likelihood maps. Note the when the target is occluded the bounding box turns red indicating low confidence mode.

Car Scale



Please note this is academic code, USE AT YOUR OWN RISK !

Code is distributed under the GNU GPL license for more information go to http://www.gnu.org/copyleft/gpl.html

This distribution includes a re-distribution of two additional code packages:
1) VL-feat 0.9.18 http://www.vlfeat.org/
2) Pioter Dollar Matlab toolbox http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html


The vehicle dataset used in this paper can be downloaded from the Real Time Tracking-With-Detection project page.

The online tracking benchmark of Wu et al. (CVPR 2013) can be found at Online Object Tracking: A Benchmark project page.


  • Presentation given at a seminar in Tel-Aviv University PPT
    Embedded videos might not play in Powerpoint versions prior to 2010


ECCV 2014 poster PDF