Andrew Price has a complete tutorial on this and I have done some work outside blender on creating a 3D scene by joining the sequences in video as well as averaging a series of frames and then pulling out moving objects by the difference as well as gradient vector flow.
This one is from U tube as a test to see if it embeds.
Interesting stuff and I am sure it will get better.
This is interesting as it is possible to "green screen" any background by image averaging and gradient vectors as well as blobbing.
I have been experimenting with taking educational videos and removing the background as a panorama backdrop, extracting the exhibitor or instructor, and then using SVG like changes to the scene to replace the instructor with an animated character or another extracted instructor, so I can change instructors and keep the same course content. Now if I could only develop a soda pop that had courses in quantum physics that you could just buy for a buck and drink the knowledge. That would be odd. "I will have a can of astrophysics, some neurosurgery ding dongs and a licorice vector algebra, and oh , uhmm, a bottle of milk with French and Chinese languages." Joking, but, you never know what the future will bring.
I purposely selected videos that do not require flash, as it is a dying medium that was never good and hopefully it will vanish completely soon.