Mining frequent tree patterns is an important problem, since tree structures are used in various fields such as computational biology, XML databases, and so on. However, mining all frequent subtrees is sometimes infeasible because of the combinatorial explosion. In this paper, by combining an efficient algorithm for enumerating free trees and the pruning techniques for mining closed and maximal rooted trees, we present an efficient algorithm named CMFree-TreeMiner that discovers only closed and maximal frequent induced free subtrees in a database of labeled free trees. Experimental results show the effectiveness of the proposed algorithm