Your question says nothing about your problem. It depends on your topic and its formulation and programming proficiency.

If it was that easy everyone would have one. Find a method you find interesting, ask about how to extend it to a new problem/decrease the error, etc. See what people have done before. Read enough papers and you will go "has anyone ever done this before?" and if no one has, that's your project.

For example, say you find 3D interpolation schemes interesting. You go, "well if the domain is square, can we decrease the error even faster?" Say you find a paper that describes how to do this. You may ask "will this generalize to 4D?" or "Can I improve this result to work for _______?". 

So, start reading the primary literature that interests you. That will give you ideas.