Image reconstruction for spatially variant systems using orthonormal natural pixels

Summary form only given. A reconstruction method which avoids the unnatural assumptions of imposed pixel geometries in most reconstruction strategies has been developed. The model incorporates detailed knowledge about the tomograph system response function and statistical properties for the acquired signal. Estimation including least-squares and maximum likelihood use in orthonormal pixel basis decomposition of the unknown spatial distribution to eliminate systematic error and minimize statistical errors that occur using square of polar pixel bases. Symmetries in the data acquisition process lead to very efficient implementations of the new inverse problem algorithms