Combined shearlet and TV regularization in sparse-view CT reconstruction

Preclinical in vivo micro computerized tomography suffers from high image noise, due to limitations on total scanning time and the small pixel sizes. A lot of different noise minimization algorithms have already been proposed to reconstruct images acquired in low dose settings. Sparse-view reconstruction amongst others can reduce acquisition dose significantly, by acquiring only a small subset of projection views. Total Variation minimization has been used extensively to solve these problems. However, the performance of TV is suboptimal for complex images, compared to simple images with little texture. This is mainly due to the underlying piecewise constant image model imposed by TV. A recent efficient solver was developed for convex problems, able to incorporate regularization terms different from TV. The work presented here is a proof-of-concept study combining both TV as well as shearlets as regularization terms into one general CT reconstruction algorithm. Shearlets, closely related to wavelets, take edges into account in a multitude of directions at different scales, and have good compaction properties. This makes shearlets a better candidate than TV for compressed sensing problems. The resulting reconstructions were compared to TV minimization and to shearlet minimization. The combination of both shows benefits for sparse-view CT imaging, and leads to edge-preserved image denoising. Difference images show a very small loss in resolution, which may be caused by difficult parameter selection.