SNR-weighted ART applied to transmission tomography

In this work, a modification of the well-known algebraic reconstruction technique (ART) is applied to the reconstruction of images based on X-ray transmission measurements. The modified version takes the noise statistics of the measured data into account. The difference between a measured and an estimated line integral that is back-projected during the iteration is weighted by a confidence measure. This measure is inversely proportional to an estimate of the signal-to-noise ratio (SNR) of the line integral. The estimation is done using Gaussian error propagation of the known uncertainty of the measured intensity. The resulting algorithm is as efficient as the original ART and shows significantly better image quality.