Super-Resolution and Sparse View CT Reconstruction

Abstract

We present a flexible framework for robust computed tomography (CT) reconstruction with a specific emphasis on recovering thin 1D and 2D manifolds embedded in 3D volumes. To reconstruct such structures at resolutions below the Nyquist limit of the CT image sensor, we devise a new 3D structure tensor prior, which can be incorporated as a regularizer into more traditional proximal optimization methods for CT reconstruction. As a second, smaller contribution, we also show that when using such a proximal reconstruction framework, it is beneficial to employ the Simultaneous Algebraic Reconstruction Technique (SART) instead of the commonly used Conjugate Gradient (CG) method in the solution of the data term proximal operator. We show empirically that CG often does not converge to the global optimum for tomography problem even though the underlying problem is convex. We demonstrate that using SART provides better reconstruction results in sparse-view settings using fewer projection images. We provide extensive experimental results for both contributions on both simulated and real data. Moreover, our code will also be made publicly available.

TomoFluid: Reconstructing Dynamic Fluid From Sparse View Videos

“…Baselines: In the following, we compare our approach to four baseline reconstruction techniques. The first baseline is an iterative tomographic reconstruction method named Simultaneous Algebraic Reconstruction Technique (SART) [3,49], since it produces reasonably high-quality results while still being applicable to arbitrary camera models and application scenarios in practice [21,26,51]. The second baseline is the Bregman algorithm of Goldstein and Osher for TV-regularized denoising (Getreuer) [18].…”

“…It involves the data-fitting term, the novel view regularizer, the reprojection consistency prior, the spatial and temporal regularizers of the density field (two first components of L smooth ) and the density consistency prior. We follow the work [51] and use the PSART algorithm as solver to tackle the proximal operators of L data and L novel efficiently in a matrix-free manner.…”

High resolution 3D visualization of the spinal cord in a post-mortem murine model

“…It should be noted that some regions of the sample in the regSIRT reconstruction look patchy and oversmoothed that worsen the final spatial resolution. Different strategies have been proposed to reduce the undesired artifacts in images with sparse projections, one of the most interesting and promising of them is super resolution technique [45].…”

Warp-and-project tomography for rapidly deforming objects

“…The main interest of using X-ray CT in computer graphics is to capture the internal structures of opaque objects [Anirudh et al 2018;Zhao et al 2011], as well as to retrieve complicated surfaces with occlusions like flowers [Ijiri et al 2014;Stuppy et al 2003;Zang et al 2018a].…”