RECONSTRUCTION OF OFF-AXIS ELECTRON HOLOGRAMS USING A NEURAL-NET

In off-axis electron holography, the information about the specimen co ntained in amplitude and phase of the image wave is recorded by superi mposing a reference wave to the image wave and recording the resulting interference pattern, called a hologram, The image wave is usually re trieved from this hologram by a filtering process in Fourier space. An alternative approach is to calculate the image wave pixel by pixel fr om the intensities in the hologram. In this paper, we describe a neura l network that performs this step, As neural networks can represent no nlinear relations between the input and output data, this reconstructi on process can use the nonlinear part of the information contained in the hologram and is thus less sensitive to noise.