An optimisation of Gaussian mixture models for integer processing units

  • Claudio Salvadori
  • Matteo Petracca
  • Jesus Martinez del Rincon
  • Sergio A. Velastin
  • Dimitrios Makris
Original Research Paper

DOI: 10.1007/s11554-014-0402-5

Cite this article as:
Salvadori, C., Petracca, M., del Rincon, J.M. et al. J Real-Time Image Proc (2014). doi:10.1007/s11554-014-0402-5

Abstract

This paper investigates sub-integer implementations of the adaptive Gaussian mixture model (GMM) for background/foreground segmentation to allow the deployment of the method on low cost/low power processors that lack Floating Point Unit. We propose two novel integer computer arithmetic techniques to update Gaussian parameters. Specifically, the mean value and the variance of each Gaussian are updated by a redefined and generalized “round” operation that emulates the original updating rules for a large set of learning rates. Weights are represented by counters that are updated following stochastic rules to allow a wider range of learning rates and the weight trend is approximated by a line or a staircase. We demonstrate that the memory footprint and computational cost of GMM are significantly reduced, without significantly affecting the performance of background/foreground segmentation.

Keywords

Gaussian mixture model Smart-camera Computer arithmetic and integer implementation Computer vision optimisation for microcontrollers Embedded systems 

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Claudio Salvadori
    • 1
  • Matteo Petracca
    • 2
  • Jesus Martinez del Rincon
    • 3
  • Sergio A. Velastin
    • 4
  • Dimitrios Makris
    • 5
  1. 1.TeCIP InstituteScuola Superiore Sant’AnnaPisaItaly
  2. 2.National Inter-University Consortium for TelecommunicationsPisaItaly
  3. 3.The Institute of Electronics, Communications and Information Technology (ECIT)Queens University of BelfastBelfastUK
  4. 4.Department of Informatic EngineeringUniversidad de Santiago de ChileSantiago de ChileChile
  5. 5.Digital Imaging Research CentreKingston UniversityLondonUK

Personalised recommendations