Figure 1.

Schematic layout of the RESIRE algorithm. RESIRE first calculates the forward projections of an arbitrary initial object using FST or the Radon transform. The differences between the calculated and measured projections are back-projected to yield the gradient. The initial object is then updated using the gradient. Optional constraints such as positivity and support can be enforced for better convergence and accuracy. The updated object is used for the next iteration. Usually, after several hundreds of iterations, the algorithm converges.