Iterative prior-knowledge-based image reconstruction algorithms

Image reconstruction from limited number of sampled data typically leads to infinitely many data-consistent solutions, and it is usually difficult to specify an accurate image even in the absence of noise. One solution to this non-uniqueness problem is to incorporate prior knowledge into the estimation procedure, and the prior discrete Fourier transform (PDFT) algorithm is a good example based on this aspect. The PDFT can promisingly provide a high-quality solution by solving the best approximation in a user-designed Hilbert space as the image estimate, while its computational cost with large data sets is unfavorable. In our previous research achievements, the implementation of the iterative algebraic reconstruction techniques (ART) and conjugate gradient (CG) method into the PDFT successfully improves computational efficiency significantly, without degrading the image quality. In this paper, some other well-known iterative algorithms were implemented in the PDFT and their performances were investigated.