Fast Image Reconstruction for Generalized Projection Imaging

Gerrit Schultz¹, Daniel Gallichan¹, Marco Reisert¹, Maxim Zaitsev¹, Jrgen Hennig¹

Recent approaches in signal localization rely on encoding with strongly nonlinear and even ambiguous field geometries. When more than two encoding fields are involved, image reconstruction becomes highly time-consuming because the encoding matrix does in general not have a sparse structure. However, for a large class of sampling trajectories, the imaging process can intuitively be interpreted as sparse image projections. In this abstract, we show that the finite duration of the acquisitions destroys the sparsity. However, with some loss of resolution along frequency-encoding the encoding matrix can be sparsified, which speeds up the reconstruction by up to two orders of magnitude.