Abstract

Tomographic data are inevitably corrupted by noise, and the number of projections available is often small. Such data cannot define an image uniquely, but are consistent with a whole range of "feasible images". Recognising that our choice of a single image is not unique, the Maximum Entropy Method chooses the feasible image which has the greatest configurational entropy: Where p_i is the proportion of intensity originating from pixel "i" and m_i is the corresponding measure or initial estimate.

© 1984 Optical Society of America