Yes, the method I suggested is simple but not perfect. When needed, I add a state-machine in order to ignore any reversal movement if the wheel does not go forward to the next phase (signals switching) and vice-versa. But this requires polling both signals fast enough to detect every state-change.
I use to say that this problem with encoders is the equivalent, though different, to rebounds in switches.

Some manufacturers make encoders with 'movement' and 'direction' outputs, instead of the conventional A and B phases.

Fran: Neither Torrecedeira nor CUVI, at the E.T.S.I. Industriales.

Regards,
Alfredo del Río.