Background Image Synthesis in a Crowded Street Scene
Using the Stauffer-Grimson algorithm [1], not only the foreground mask of the moving object can be extracted, but we may also generate a high-quality background image, even if the street is never 'empty' during the observation period.
Left: a video frame. Right: synthesised background image
Demo video:
The following demo video demonstrates how the background image is created from an arbitrary initial position, and how is updated, following the changes of the scene (e.g. a new parking car appears):
Reference:
[1] C. Stauffer and W. E. L. Grimson, “Adaptive background mixture models for real-time tracking,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition, vol. 1, pp. 22–29, 1999.
