A geometric projection-space reconstruction algorithm (Q2639625)

The authors present a method to reconstruct images from finite sets of noisy projections (parallel geometry) that may be available only over limited or sparse angles. The algorithm is based on a data completion procedure before applying the standard convolution backprojection algorithm. To this end a stochastical model of the tomographic problem is developed. The algorithm calculates the maximum a posteriori estimate of the full sinogram from the available data. A priori knowledge of the convex hull of the object is incorporated and consistency conditions of low order are exploited. The method is illustrated with reconstructions from simulated data.