SPATIAL-RESOLUTION PROPERTIES OF PENALIZED-LIKELIHOOD IMAGE-RECONSTRUCTION - SPACE-INVARIANT TOMOGRAPHS

This paper examines the spatial resolution properties of penalized-lik elihood image reconstruction methods by analyzing the local impulse re sponse. The analysis shows that standard regularization penalties indu ce space-variant local impulse response functions, even for space-inva riant tomographic systems, Paradoxically, for emission image reconstru ction, the local resolution is generally poorest in high-count regions . We show that the linearized local impulse response induced by quadra tic roughness penalties depends on the object only through its project ions. This analysis leads naturally to a modified regularization penal ty that yields reconstructed images with nearly uniform resolution. Th e modified penalty also provides a very practical method for choosing the regularization parameter to obtain a specified resolution in image s reconstructed by penalized-likelihood methods.