TomoFluid: Reconstructing Dynamic Fluid From Sparse View Videos

Abstract

Visible light tomography is a promising and increasingly popular technique for fluid imaging. However, the use of a sparse number of viewpoints in the capturing setups makes the reconstruction of fluid flows very challenging. In this paper, we present a state-of-the-art 4D tomographic reconstruction framework that integrates several regularizers into a multi-scale matrix free optimization algorithm. In addition to existing regularizers, we propose two new regularizers for improved results: a regularizer based on view interpolation of projected images and a regularizer to encourage reprojection consistency. We demonstrate our method with extensive experiments on both simulated and real data.

IntraTomo: Self-supervised Learning-based Tomography via Sinogram Synthesis and Prediction

“…FDK in 3D), since it is still the most popular tomographic reconstruction algorithm [52]. The second baseline is SART, a robust iterative reconstruction algorithm [38,3], which is flexible and can be adapted to different camera models and applications [29,32,70]. A TV regularized algorithm [27] is used as the third baseline (TV).…”

“…The computed tomography technique is often associated with medical X-ray tomography. However, it has numerous other applications at different scales and involving various techniques to produce the projections such as: industrial X-ray CT [17,69], synchrotron X-ray tomographic microscopy [19,56], electron tomography microscopy [46] and visible light tomography [31,70]. Tomography techniques operate on a set of projections of the scanned object from different angles.…”

NeAT: Neural Adaptive Tomography

“…While most CT imaging is performed with X-rays due to their ability to penetrate a wide range of materials [Kak and Slaney 2001], there have also been a number of works utilizing visible light, especially in the visual computing community (e.g. [Atcheson et al 2008;Eckert et al 2019;Gregson et al 2012;Hasinoff and Kutulakos 2007;Zang et al 2020]).…”

Spatiotemporal tomography based on scattered multiangular signals and its application for resolving evolving clouds using moving platforms

“…In nearly all CT approaches, the object is considered static during image acquisition. However, in many cases of interest, the object changes while multi-view images are acquired sequentially [11,53]. Thus, effort has been invested to generalize 3D CT to four-dimensional (4D) spatiotemporal CT, particularly in the computer vision and graphics communities [42,52,53].…”

“…However, in many cases of interest, the object changes while multi-view images are acquired sequentially [11,53]. Thus, effort has been invested to generalize 3D CT to four-dimensional (4D) spatiotemporal CT, particularly in the computer vision and graphics communities [42,52,53]. This effort has been directed at linear-CT modalities.…”