The main problem addressed by this dissertation is the accurate and automated determination of electron microscope imaging conditions. This enables the restoration of the object wave, which confers direct structural information about the specimen, from sets of differently aberrated images. An important member in the imaging chain is the image recording device, in many cases now a charge-coupled device (CCD) camera. Previous characterisations of these cameras often relied on the unjustified assumption that the Modulation Transfer Function (MTF) also correctly describes the spatial frequency dependent attenuation of the electron shot noise. A new theory is therefore presented that distinguishes between signal and noise transfer. This facilitates the evaluation of both properties using a detailed Monte-Carlo simulation model for the electron and photon scattering in the scintillator of the camera. Furthermore, methods for the accurate experimental determination of the signal and noise transfer functions are presented. In agreement with the Monte-Carlo simulations, experimental results for commercially available CCD cameras show that the signal transfer is significantly poorer than the noise transfer. The centrepiece of this dissertation is the development of new methods for determining the relative aberrations in a set of images and the absolute symmetric aberrations in the restored wave. Both are based on the analysis of the phase information in the Fourier domain and give each Fourier component a weight independent of its strength. This makes the method suitable even for largely crystalline samples with little amorphous contamination, where conventional methods, such as automated diffractogram fitting, usually fail. The method is then extended to also cover the antisymmetric aberrations, using combined beam tilt and focal series. The applicability of the new method is demonstrated with object wave restorations from tilt and focal series of complex inorganic block oxides and of carbon nanotubes filled with one-dimensional inorganic crystals. The latter specimens allowed for the first time a direct comparison between the phase shift in the restored object wave of a specimen with precisely known thickness and the value predicted by simulations.

The main problem addressed by this dissertation is the accurate and automated determination of electron microscope imaging conditions. This enables the restoration of the object wave, which confers direct structural information about the specimen, from sets of differently aberrated images. An important member in the imaging chain is the image recording device, in many cases now a charge-coupled device (CCD) camera. Previous characterisations of these cameras often relied on the unjustified assumption that the Modulation Transfer Function (MTF) also correctly describes the spatial frequency dependent attenuation of the electron shot noise. A new theory is therefore presented that distinguishes between signal and noise transfer. This facilitates the evaluation of both properties using a detailed Monte-Carlo simulation model for the electron and photon scattering in the scintillator of the camera. Furthermore, methods for the accurate experimental determination of the signal and noise transfer functions are presented. In agreement with the Monte-Carlo simulations, experimental results for commercially available CCD cameras show that the signal transfer is significantly poorer than the noise transfer. The centrepiece of this dissertation is the development of new methods for determining the relative aberrations in a set of images and the absolute symmetric aberrations in the restored wave. Both are based on the analysis of the phase information in the Fourier domain and give each Fourier component a weight independent of its strength. This makes the method suitable even for largely crystalline samples with little amorphous contamination, where conventional methods, such as automated diffractogram fitting, usually fail. The method is then extended to also cover the antisymmetric aberrations, using combined beam tilt and focal series. The applicability of the new method is demonstrated with object wave restorations from tilt and focal series of complex inorganic block oxides and of carbon nanotubes filled with one-dimensional inorganic crystals. The latter specimens allowed for the first time a direct comparison between the phase shift in the restored object wave of a specimen with precisely known thickness and the value predicted by simulations.