Learning Efficient Linear Predictors for Motion Estimation


Matas, Jiří, Zimmermann, Karel, Svoboda, Tomáš, Hilton, Adrian
ICVGIP, Berlin,Germany
Proceedings of 5th Indian Conference on Computer Vision, Graphics and Image Processing
Pages 445-456
December, 2006

Abstract

A novel object representation for tracking is proposed. The tracked object is represented as a constellation of spatially localised linear predictors which are learned on a single training image. In the learning stage, sets of pixels whose intensities allow for optimal least square predictions of the transformations are selected as a support of the linear predictor. The approach comprises three contributions: learning object specific linear predictors, explicitly dealing with the predictor precision -- computational complexity trade-off and selecting a view-specific set of predictors suitable for global object motion estimate. Robustness to occlusion is achieved by RANSAC procedure. The learned tracker is very efficient, achieving frame rate generally higher than 30 frames per second despite the Matlab implementation.

Keywords

tracking, real-time, motion estimation

Full Paper

Portable document format file PDF
Gzipped postscript file ps.gz ()

Bibtex entry

@InProceedings{Zimmermann-ICVGIP-2006,
  author = 	 {Matas, Ji{\vr}{\'i} and Zimmermann, Karel and 
                  Svoboda, Tom{\'a}{\vs} and Hilton, Adrian},
  title = 	 {Learning Efficient Linear Predictors for Motion Estimation},
  booktitle =    {Proceedings of 5th Indian Conference on Computer Vision, 
                  Graphics and Image Processing},
  publisher =    {Springer-Verlag},
  address = 	 {Berlin,Germany},
  issn =	 {0302-9743},
  isbn =	 {978-3-540-68301-8},
  series =       {LNCS4338},
  year = 	 {2006},
  pages =        {445-456},
  book_pages =   {965},
  venue =  	 {Madurai, India},
  month = 	 {December},
  day =          {13-16},
  editor =       {Rangachar Kasturi, Subhashis Banerjee},
  organization = {Thiagarajar College of Engineering},
  psurl =        {[zimmerk-icvgip06.pdf]},
  keywords =     {tracking, real-time, motion estimation},
  annote = {A novel object representation for tracking is proposed.
    The tracked object is represented as a constellation of spatially
    localised linear predictors which are learned on a single training
    image.  In the learning stage, sets of pixels whose intensities
    allow for optimal least square predictions of the transformations
    are selected as a support of the linear predictor.  The approach
    comprises three contributions: learning object specific linear
    predictors, explicitly dealing with the predictor precision --
    computational complexity trade-off and selecting a view-specific
    set of predictors suitable for global object motion estimate.
    Robustness to occlusion is achieved by RANSAC procedure.  The
    learned tracker is very efficient, achieving frame rate generally
    higher than 30 frames per second despite the Matlab
    implementation.},
}