Multiple Dimension Chrominance Model for Background Subtraction

B.R. Williams and M. Zhang (USA)


Chrominance Model, Illumination-invariant, Background Subtraction * This research is supported by Christopher Newport University Applied Research Center Funding and Dean’s Grant of College of Liberal Arts and Sciences, Christopher Newport University.


Background subtraction is a very popular technique for image processing in video surveillance systems. Current background subtraction techniques work well under constant lighting conditions. However, changes in lighting (from clouds, time of day, etc) disturb most methods. Techniques have been proposed to respond to lighting changes, but regional lighting changes and changing background scenery have been found to be problematic. In this paper a Multiple Dimension Chrominance Model (MDCM) is proposed that builds upon an earlier chrominance model and handles both regional light changes and changes in background scenery. The technique divides pixels into their YUV color components and analyzes the components individually. Adaptations are made for very dark pixels with little chrominance information. Final results show that, in our case, this model outperforms the model it was derived from by 2.77%.

Important Links:

Go Back