Reduction of metal streak artifacts in X-ray computed tomography using a transmission maximum a posteriori algorithm

A review of automated image understanding within 3D baggage computed tomography security screening

“…The vast majority of MAR-based CT literature is found in the medical domain and techniques typically fall into one of three categories: 1) sinogram (or projection) completion methods, whereby artefact correction is performed by replacing corrupted projection data prior to CT reconstruction [50,69,115,74]; 2) iterative methods, whereby iterative reconstruction techniques are used to generate superior quality reconstructions [107,39,92] and 3) hybrid methods, using combinations of (1) and (2) [58,13,61]. While many of these published techniques claim substantial improvements to previous methods, these claims are often based on limited comparisons.…”

“…Despite the development of optimised approaches such as Ordered Subset Expectation Maximisation (OSEM) [46], the Row-Action Maximum Likelihood Algorithm (RAMLA) [16], Model-Based Iterative Reconstruction (MBIR) approaches [113], Iterative Coordinate Descent (ICD) optimisation [14,95], Block-Iterative (BI) modifications [18] and numerous hybrid methods [58,13,61], high computational cost remains the major factor preventing the universal implementation of such techniques in commercial CT machines (even in the medical domain where the demand for high throughput are not paramount).…”

Discrete Tomography Reconstruction Algorithms for Images with a Blocking Component

“…This is a common problem in many tomographical applications that may cause artifacts in the rebuilt images, that in some cases may reduce the quality of the reconstruction up to the point of making the resulting image unusable, for example for diagnostic purposes in the case of a medical scan. Many heuristics exists to reduce such artifacts, based on algebraic approaches [20], statistical analysis [8], linear interpolation [17], partial differential equations [9], and image impainting [11]. The described approaches are often sufficient in applications, however to the best of our knowledge there is no theoretical study that determines the basic properties of consistency and uniqueness of these problems.…”

Decoupling the equations of regularized tomography

“…1) [10,11,12]. Since the actual projection data and scanner parameters were unavailable, we simulated the projections (using Matlab) after rescaling the image pixel values (range 0 to 255) by 0.012 (Fig.…”

“…In addition, we assume the idealized line-integral model of projection, while discrete regularization techniques more accurately model real scanners [12]. Nonetheless, we can still demonstrate ( §4) that our method can remove noise.…”