Technically speaken a video is not much else then a lot if images (frames) in sequence. So there is not much difference between "learning" from an image then from a video (despite some more information you may obtain from the video).
Your question however cannot be answered easily. If you ask AI enthusiasts they would probably answer of course it's possible to "teach" a machine from a video. However this is not an easy task at all. In fact it is a topic of worldwide research which several million (billion) of dollars are spend on and many researchers are investing their (life)time to make some steps towards a self learning machine. This is no goal which can be achieved during an afternoons work, but for some it is an aim which they dedicate their whole life too.
If you really want to work in this field you should be absolutely clear about the challenge this implies. You can try to do this on your own from pure internet / book learning but if possible in any way in my opinion, you should consider taking courses / studies in the field of pattern learning, machine learing, articifial intelligence or similar.
For the start there are many ressources online which you can find fairly easy with the keywords: "pattern recognition", "machine learning", "computer vision" in combination with tutorial or excercise, examples and so on. Just some of them:
-
http://www.r-bloggers.com/in-depth-introduction-to-machine-learning-in-15-hours-of-expert-videos/[
^]
http://fastml.com/machine-learning-courses-online/[
^]
You could as well take part in online courses, that are offered e.g.
on
Coursera[
^]. They are usually free and open to public access.
You can also use google scholar to find scientific publications related to the topic, which are the key ressource if you want to find out which progress is made in this field recently:
http://scholar.google.de/scholar?hl=de&q=machine+learning&btnG=&lr=[
^]