KEYWORDS: Image restoration, Education and training, X-ray imaging, Deep learning, X-rays, Reconstruction algorithms, Image quality, Data modeling, Point spread functions, Algorithm development
A novel method for image restoration is introduced that uses a synthetic prior intermediate (SPI) which is passed through a forward imaging operator, creating a data pair well-structured for inverse operator optimization, of which network training is of particular interest. This technique is applied to a critical problem in x-ray reconstruction: noise and artefact removal. We discuss the creation of the SPI through state-of-the-art Deep Learning Reconstruction (DLR), a spatially variant heuristic data-driven forward model for spectrally accurate noise and artefact modelling, and final image reconstruction via a convolutional neural network. Qualitative and quantitative performance is then benchmarked on a range of samples, comparing legacy reconstruction (FDK), state-of-the-art DLR, and SPI based reconstruction. SPI based reconstruction better recovers small features while also reducing residual sampling artefacts in large features. Quantitative analysis of SPI reconstruction showed a 40% throughput improvement relative to the state-of-the-art at a comparable image quality.
X-ray computed laminographic tomography (CLT) is a viable tool for creating high-throughput volumetric imaging of large, planar samples. In this work, we present a self-supervised deep image restoration workflow to produce noise-free, artifact-free volumetric reconstructions for laminographic tomography. We demonstrate our CLT method on a variety of samples scanned with an in-house prototype system, showing that our proposed method notably outperforms classic reconstruction methods, that has the potential for more accurate detection of defects and estimation of critical dimensions, thereby providing a feasible solution for rapid inline inspection and failure analysis in advanced integrated circuits packaging.
In this work we consider a well-known empirical beam-hardening correction algorithm applied to metal artifact reduction in cone beam micro-computed tomography (CT). In its basic form, this algorithm consists of a few simple steps: segmentation of metal components out of uncorrected image data, forward projection to create metal only projection data, and the creation of several correction basis images. The set of basis images can then be combined to create a corrected image with mitigated metal artifacts. The combination weights can be determined either manually or automatically by solving an optimization equation, be performed globally or locally (allowing weights to vary spatially). The latter approach may be more attractive as it may account for larger local variances in scatter artifacts, however, not practical for manual optimization, requiring an automated approach. We apply both global and spatially variant version of the algorithm to datasets from a cone-beam micro-CT (using a 3D X-ray microscope) and study the performance.
A novel automated workflow for the recovery of image resolution using deep convolutional neural networks (CNNs) trained using spatially registered multiscale data is presented. Spatial priors, coupled with high order voxel-based image registration, are used to correct for uncertainties in image magnification and position. A network is then trained to remove the effects of point spread from the low-resolution data, improving resolution while reducing image noise and artefact levels. While benchmarking on real materials, including biological, materials science and electronics samples, we find that resolution recovery improves quantitative and qualitative measurements, even if certain image details cannot be easily identified from the original low-resolution data.
Metal artifacts are one of the most common reasons for reduced image quality and usability in polychromatic cone-beam CT. In this work, we revisit empirical beam hardening correction algorithm and propose a few practical optimizations to simplify its application. First, fuzzy C-means segmentation method is used to perform an automatic segmentation of the metal component. Second, a minimum variance optimization technique provides a suitable combination of correction basis images. Finally, a sub-volume (spatially varying) optimization method is used to account for a varying contribution of metal artifacts through the image. We apply the modified algorithm to datasets from cone-beam CT and evaluate its performance.
A novel automated workflow for the recovery of image resolution using deep convolutional neural networks (CNNs) trained using spatially registered multiscale data is presented. Spatial priors, coupled with high order voxel-based image registration, are used to correct for uncertainties in image magnification and position. A network is then trained to remove the effects of point spread from the low-resolution data, improving resolution while reducing image noise & artefact levels. While benchmarking on real materials, including biological, materials science and electronics samples, we find that resolution recovery improves quantitative and qualitative measurements, even if certain image details cannot be easily identified from the original low-resolution data.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.