Please do help me how do I read the encoder correctly.
Again (if you had searced, you would have found this):
A as clock and b equals direction does not cut it.
This is an old 'simple' design that work very well if fed to a hardware up/down counter and you you can live with an error of a few counts per rev.
1) there is one of 8 (or is it one of 16) direction changes that get lost.
2) when the encoder sit just over a border, you can get 'steps' with a frequency of several megahertz.

two, maybe 3 solutions work
The agilent chip
a CPLD programmed to react very fast and consider direction based on previous and current state.
One, I have not tried, comes to mind: Instead of a CPLD use a 100MHZ Cygnal processor, it might keep up.

Erik