PROCESS FOR ESTIMATING AN IMAGE FUNCTION BY ITERATIONS

摘要

A description is given of a process for the iterative estimation of an image function f from a series of measurements m using a conjugate gradient type algorithm permitting the reversal of an equation of type Af=m. According to the invention, the aim is not to symmetrize the matrix A before applying the process, which has hitherto been considered necessary in order to ensure the convergence of the algorithm and instead the aim is to make the system of measuring equations quasi-symmetrical by applying an appropriate preconditioning operator. It is then possible to apply an iterative reconstruction algorithm aiming at minimizing the standard of the residue and where the updating of the descent directions takes place in accordance with the residues and not the gradients. The convergence of the estimate is then saved and is much faster.