Unreliable failure detectors are mechanisms providing information about process failures. This paper addresses the implementation of the Eventually Perfect failure detector class, denoted P, in the crash-recovery failure model. The approach followed is to transform existing implementations of the Omega failure detector, which provides an eventual leader election functionality, into P. In a system with the minimal requirements to implement P, we first propose a generic transformation algorithm that uses the underlying Omega as a black box. Then, we enhance the algorithm to provide a common set of k correct processes. The algorithms do not require the membership of the system to be known a priori.