No abstract

No abstract

No abstract

No abstract

The reconstruction of an image from incomplete view data requires the use of several constraints not derived from ray sum (projection data) measurements. The constraints can be incorporated through the method of (sequential and parallel) projections onto the constraint sets. These methods for the use of information regarding the noise and the image are implemented and compared in this paper. It is shown that the use of noise statistics decreases the mean square error in the image and that the method of parallel projections results in smaller error than the method of sequential projections if a sufficient number of iterations is permitted.