I struggle with this question too. It's always tough since as the creator you always see the flaws in what you're building. Here are my 2c, hope it helps:
1. What is your core value prop? An MVP isn't a set of features necessarily, but it should solve a specific problem that you think your users have.
2. What do you still need to learn about your users? What is the shortest path to learning those things? If your product is like most others, it will almost certainly change once you get it into the hands of users. Optimize for learning, not "correctness".
3. What is stopping you from launching? For me, it's often a fear of failure. When it isn't, I rarely question whether I should launch or not. I know the product isn't ready. I then ask myself, what am I afraid of? Usually it's that either 1) no one will care or 2) people will care and the product won't be ready so people will churn.
Taking those individually, I try to shift "no one caring" to actually relieve pressure. If no one cares, then what's there to be afraid of? As for 2, if the product isn't ready and people churn, there will be more users and you will learn a ton.
I find it's helpful to think of launching an MVP as a learning opportunity. I mean this in the most genuine, helpful way: the most likely case is that people won't care and there is a ton of learning in that.
One last thing: you may have already read this but PG has an essay called The 18 Mistakes That Kill Startups that covers both launching too early and launching too late: http://www.paulgraham.com/startupmistakes.html
I hope this was helpful, best of luck!