A fast recursive algorithm for computing CR-type bounds for image reconstruction problems

The authors describe a method for computing matrix CR bounds for image reconstruction problems using an iterative algorithm that avoids the intractable inversion of the Fisher matrix required by direct methods. The algorithm produces a close approximation to the CR bound, requiring only O(n/sup 2/) floating point operations per pixel of interest, an order of magnitude savings relative to the O(n/sup 3/) flops required by noniterative methods. To illustrate the utility of the iterative algorithm, a prototypical application is studied: the dependence of achievable reconstruction accuracy on angular and radial sampling.<
>