After the first review process, could the rebuttal totally change the reviewers' minds, i.e. from accept to reject or the other way around?
After the first review process, could the rebuttal totally change the reviewers' minds
Very likely no. The rebuttal letter is usually very short (a few hundred words, typically) in comparison to the actual paper, so it is unlikely that the few extra explanations will turn the reviewer around entirely. Also, what is true for most humans is also true for academics - once people have formed their opinion about something, it is hard to fundamentally change their mind. That is, if a reviewer hates a paper enough to vote for full reject in the first round, a few good explanations will not be enough to turn her/him around to accept.
from accept to reject
Not unless you write something really idiotic ("I don't understand the criticism that this paper is drawing. I am sure, Prof. Bigshot did not have to listen to such criticism when he originally published the same contribution."). Also, if your paper is accepted, what is there to rebute?
or the other way around?
As jakebeal writes, the most likely case is that the rebuttal letters are used to differentiate between a number of borderline papers for a few remaining conference slots. In my (limited) experience in the matter, nobody will even really look at rebuttal letters for papers that are already clearly accepted or rejected.
First, it's worth noting that very few computer science conferences actually allow a rebuttal: mostly, the decision you get is the final decision. In those rare cases that I have encountered that have a rebuttal period, it mainly serves to help disambiguate papers that are near the borderline. Thus, it can certainly shift the status of a paper between accept and reject (e.g., by showing that reviewer #3 really was nuts and should be discounted or by eliminating the doubt that the authors were benefiting from), but is unlikely to do so unless the paper was already quite close to the boundary.