Researchers have adapted generative models, specifically Wasserstein GAN with gradient penalty (WGAN-GP) and Pix2Pix, for interpolating gaps in photon-counting X-ray imaging detectors. These models were trained on nanoCT projections with simulated gaps and compared against traditional methods like linear interpolation and Navier--Stokes inpainting, as well as a deep inpainting network called MADF. The generative models demonstrated superior performance in perceptual similarity and histogram distance, with Pix2Pix requiring less training time and producing more natural gap filling when applied to sinograms, though the improvement in reconstruction quality was marginal. AI
IMPACT This research demonstrates a novel application of generative AI for improving image quality in scientific imaging, potentially enhancing downstream analysis in fields like medical imaging.
RANK_REASON Academic paper detailing a novel application of generative models to a specific scientific imaging problem. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
