Distinguishing new image frame with openCV

Hey all,

So I have a general question about the best approach to tackle realtime pattern / face recognition and analysis upon a new “face/pattern” found.

to go into a bit more detail, lets say I have user1’s face. user1 is moving around, turning profile, smiling, etc etc. user1 leaves, and user2 approaches. What I want to do is when the new first frame of user2 is detected, save image. Essentially creating a real time data base of new users. This will need to also distinguish between the previous users and not create false positives.

My issue is fully understanding the best approach. I had a look at the CV library libface:

Libface link

and was just curious if this is a good starting point. some information I could gather would be distinguishing extreme differences in the histogram of the image, (and obvious installation concerns such as consistent lighting), but I feel this isnt enough to analyze this type of information.

suggestions? Cheers.