English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

X-ray micro-computed tomography (micro-CT) has been widely leveraged tocharacterise pore-scale geometry in subsurface porous rock. Recent developmentsin super resolution (SR) methods using deep learning allow the digitalenhancement of low resolution (LR) images over large spatial scales, creatingSR images comparable to the high resolution (HR) ground truth. This circumventstraditional resolution and field-of-view trade-offs. An outstanding issue isthe use of paired (registered) LR and HR data, which is often required in thetraining step of such methods but is difficult to obtain. In this work, werigorously compare two different state-of-the-art SR deep learning techniques,using both paired and unpaired data, with like-for-like ground truth data. Thefirst approach requires paired images to train a convolutional neural network(CNN) while the second approach uses unpaired images to train a generativeadversarial network (GAN). The two approaches are compared using a micro-CTcarbonate rock sample with complicated micro-porous textures. We implementedvarious image based and numerical verifications and experimental validation toquantitatively evaluate the physical accuracy and sensitivities of the twomethods. Our quantitative results show that unpaired GAN approach canreconstruct super-resolution images as precise as paired CNN method, withcomparable training times and dataset requirement. This unlocks newapplications for micro-CT image enhancement using unpaired deep learningmethods; image registration is no longer needed during the data processingstage. Decoupled images from data storage platforms can be exploited moreefficiently to train networks for SR digital rock applications. This opens up anew pathway for various applications of multi-scale flow simulation inheterogeneous porous media.

A comparative study of paired versus unpaired deep learning methods for physically enhancing digital rock image resolution