- Bootstrapping: During the training period, the background is not available in some environments. Then, it is impossible to compute a representative background image.
- Camouflage: A foreground object's pixel characteristics may be subsumed by the modeled background. Then, the foreground and the background cannot be distinguished.
- Foreground aperture: When a moved object has uniform colored regions, changes inside these regions may not be detected. Thus, the entire object may not appear as foreground.The foreground mask contains false negative detections.