The effects of finite sampling on reconstructed image resolution and noise

Summary form only given. A technique which models the relationship between reconstructed image resolution and noise by evaluating the norm of the image covariance matrix versus the number of reconstructed generalized pixel (object basis function) amplitudes has been developed. Data acquisition is represented by a linear transformation of functions on a continuous domain object space to functions on a discrete domain observation space; thus, the model is easily adapted to include physical effects found in many imaging modalities. A tomography system that has parallel sampling with equally spaced angles and bins was analyzed. These studies show a resolution threshold beyond which improvements in resolution, characterized by the number of orthonormal pixels reconstructed, are achieved only with large increases in image noise