Scholarlink

<p>Optical imaging under low-light conditions and through scattering media remains a major challenge for quantitative diagnostics, especially in particle image velocimetry (PIV) and microfluidic applications. These challenges are exacerbated in supercritical fluids, where strong thermodynamic gradients near the pseudo-boiling line induce non-linear refractive index fluctuations that degrade image quality and generate speckle patterns. While data-driven image reconstruction methods have shown promise in addressing such problems, their development is constrained by the lack of experimentally representative training data. This work presents an experimental dataset of low-light, micro-PIV-like image sequences acquired in a microchannel with flowing carbon dioxide across high-pressure, transcritical conditions. The dataset comprises paired corrupted and speckle-free ground-truth images generated using stationary fluorescent tracer particles, ensuring that observed distortions arise solely from optically induced effects. Images were recorded using a high-speed CCD camera under realistic noise conditions, and each image sequence is accompanied by thermodynamic metadata. The dataset is designed to support the development, validation, and benchmarking of image denoising, speckle reconstruction, and physics-informed learning methods under realistic experimental conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Experimental Dataset of Ground-Truth-Paired Micro-PIV Images in Variable-Density Supercritical CO2

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • Norbert Krauss,
  • Enrique Hurtán,
  • Joan Calafell,
  • Lluís Jofre

摘要

Optical imaging under low-light conditions and through scattering media remains a major challenge for quantitative diagnostics, especially in particle image velocimetry (PIV) and microfluidic applications. These challenges are exacerbated in supercritical fluids, where strong thermodynamic gradients near the pseudo-boiling line induce non-linear refractive index fluctuations that degrade image quality and generate speckle patterns. While data-driven image reconstruction methods have shown promise in addressing such problems, their development is constrained by the lack of experimentally representative training data. This work presents an experimental dataset of low-light, micro-PIV-like image sequences acquired in a microchannel with flowing carbon dioxide across high-pressure, transcritical conditions. The dataset comprises paired corrupted and speckle-free ground-truth images generated using stationary fluorescent tracer particles, ensuring that observed distortions arise solely from optically induced effects. Images were recorded using a high-speed CCD camera under realistic noise conditions, and each image sequence is accompanied by thermodynamic metadata. The dataset is designed to support the development, validation, and benchmarking of image denoising, speckle reconstruction, and physics-informed learning methods under realistic experimental conditions.