I suggest you read through this thread: Introducing new project: image2sdr.
You would need to sample images out of the video to create an SDR stream. Then create a Network with an SP and TM. Then you could get anomaly values.
What type of video are you planning on analyzing?