The implementation of an algorithm designed to improve images of the heart reconstructed from few projections is described. The concept underlying the technique is due to G. C. McKinnon and R. H. T. Bates (1981). Described are refinements to their algorithm, incorporating ideas suggested by P. B. Heffernan and R. A. Robb (1982) which allow its practical application in cardiac imaging, where the projection data may be far from ideal. The algorithm's efficacy is demonstrated using experimentally measured data. The computational cost of the procedure is assessed and found to be especially acceptable in the case of reconstruction of a three-dimensional volume comprising a stack of transverse sections. The implications of the results are discussed in the context of automatic object definition, and real-time visualization of the moving structure of the heart.