TIMBIR: A Method for Time-Space Reconstruction From Interlaced Views

Abstract

Synchrotron X-ray computed tomography (SXCT) is increasingly being used for 3D imaging of material samples at micron and finer scales. The success of these techniques has increased interest in 4D reconstruction methods that can image a sample in both space and time. However, the temporal resolution of widely used 4D reconstruction methods is severely limited by the need to acquire a very large number of views for each reconstructed 3D volume. Consequently, the temporal resolution of current methods is insufficient to observe important physical phenomena. Furthermore, measurement non-idealities also tend to introduce ring and streak artifacts into the 4D reconstructions.In this paper, we present a time-interlaced model-based iterative reconstruction (TIMBIR) method which is a synergistic combination of two innovations. The first innovation, interlaced view sampling, is a novel method of data acquisition which distributes the view angles more evenly in time. The second innovation is a 4D model-based iterative reconstruction algorithm (MBIR) which can produce time-resolved volumetric reconstruction of the sample from the interlaced views. In addition to modeling both the sensor noise statistics and the 4D object, the MBIR algorithm also reduces ring and streak artifacts by more accurately modeling the measurement non-idealities. We present reconstructions of both simulated and real X-ray synchrotron data which indicate that TIMBIR can improve temporal resolution by an order of magnitude relative to existing approaches.

TomoBank: a tomographic data repository for computational x-ray science

“…While datasets transition from 'challenge' to 'reference' catalogs we also aim to create a network through tomoBank in which computational scientists can not only use existing data but also request a new ad hoc measurement to validate a new method or vice-versa to optimize the experiment design. For some advanced tomographic algorithms, such new measurements are crucial for experimental validation: optimizing the experiment for the chosen algorithm is either required, or can lead to significantly more accurate results [2,3].…”

Four dimensional material movies: High speed phase-contrast tomography by backprojection along dynamically curved paths

“…Sophisticated regularized approaches exploit for example sparsity in the time evolution 6 , but some kind of prior information and/or special sampling are required 6–8 . To this end, projection angle sequences with large angular increments are best suited to estimate the spatiotemporal evolution 6,9 . Features of the object can then for example be triangulated from almost orthogonal directions, recorded at minimum time difference.…”

Space-time tomography for continuously deforming objects

“…Several acquisition strategies were proposed in the literature with this goal in mind. The closest to our approach is the interlaced view sampling approach by Mohan et al [2015], who proposed to divide the total number of projections N into N s f groups consisting of N θ projections each (Figure 2, right). Each group of projections is used to reconstruct a different volume, corresponding to a time frame in an animation of the deforming object.…”

“…On the other hand, N θ should be small enough to avoid the "motion blur" caused by the deformations. In other words, there is a trade-off between spatial and temporal resolution that requires parameter selection, but unlike in the work of Mohan et al [Mohan et al 2015], this parameter choice is strictly post-capture and can be informed by visually analyzing the amount of motion in the raw projection images. Rose dataset.…”