Nonlinear diffusion constraints for reconstructing subsampled rotational angiography data
Résumé
Interventional imaging with cone-beam C-arm CT often lacks sufficient sampling. Compressed sensing based reconstruction algorithms have shown promising results to improve image quality in this context using sparsity constraints. Compressed sensing theory by itself assumes random measurements and l1 penalties. In practice, benefits are seen with uniform subsampling patterns. Here, we investigate substituting l1 total variation with a nonlinear diffusion constraint and show on a clinical data set that image quality is also improved. This result adds flexibility to the design of CS-based algorithms as C-arm CT images may not be so well approximated by piecewise constant functions.