Bayesian estimation of 3-D objects from few radiographs

Radiography is an effective method for nondestructive inspection of solid objects, but more complete analysis and diagnoses are possible with full 3-D reconstructions. The quantity of data available in radiography is seldom sufficient to form a deterministic estimate. We present methods for statistically optimal reconstructions of 3-D objects from small numbers of digitized radiographs, which not only restore more complete spatial information, but also improve the detectability of aberrations and flaws in solid materials. The formulation includes scatter compensation under the statistical model.<
>