Phase retrieval is a well-established method for the recovery of an object wave from its magnitude-only measurements, with applications in fields such as material science, biology, and astronomy. In order to guarantee a stable and global solution for phase retrieval, measurement diversity is frequently used. However, this process requires taking several measurements, which hinders some of the advantages of phase retrieval compared with interferometric techniques. Here we propose a novel diversity framework that we refer to as digital resampling diversity. The proposed resampling diversity along with sparsity-constrained object reconstruction enables improved object reconstruction from a single squared magnitude image, without using object support constraint. We demonstrate this framework with simulation and experimental results.