The field of cryo-electron microscopy has made astounding advancements in the past few years, mainly due to improvements in the hardware of the microscopes. Yet, one of the key open challenges of the field remains the processing of heterogeneous data sets, produced from samples containing particles at several different conformational states. For such data sets, one must first classify their images into homogeneous groups, where each group corresponds to the same underlying structure, followed by reconstruction of a three-dimensional model from each of the homogeneous groups. This task has been proven to be extremely difficult. In this paper we present an iterative algorithm for processing heterogeneous data sets that combines the classification and reconstruction steps. We prove accuracy and stability bounds on the algorithm, and demonstrate it on simulated as well as experimental data sets.