Ah, I see the section you are talking about... it's referring to figure 3.50, not 3.51.....
Anyway, assuming no sub threshold conduction as the book states, and vin < Vth1, you see M1 is off. Let's say node X is 0, and Vb is > Vth2. You see that M2 will turn on due to channel forming under the gate. This will "pull up" on node X until that node voltage becomes around Vth2 lower then Vb. Once that happens, the channel goes away, M2 turns off and you have Vx ~= vb-vth2. It's not exactly Vb-vth2 since MOS doesn't really turn on/off that sharply.