Analytical approach to channelized Hotelling observer performance for regularized tomographic image reconstruction

Our goal is to analyze regularized image reconstruction methods such as penalized likelihood with respect to the performance of the channelized Hotelling observer (CHO) in the task of detecting a small target signal in the reconstructed images, in the presence of a correlated random background. We derive here an approximation to the performance of the CHO by working entirely with continuous-space formulations and then discretizing the final result. This approach leads to an extension and a refinement of approximations that we previously derived in the discrete space.