image: Figure 1: Schematic diagram of the CACTI system working principle
Credit: Song, Y., et al.
A study published in iOptics introduces a new system for high-speed imaging, potentially making advanced scientific visualization accessible to broader research communities.
The new system, known as Coded Aperture Compressive Temporal Imaging (CACTI), merges optical engineering with artificial intelligence (AI) to overcome traditional limitations in high-speed photography.
"Conventional high-speed imaging systems face constraints in hardware complexity, data throughput, and storage demands," explained the study’s corresponding author Professor Xin Yuan. "Our CACTI framework optically encodes 20 sequential frames into a single measurement and reconstructing videos at 1200 fps through deep learning algorithms, achieving data reduction while maintaining exceptional fidelity."
The system's employs a digital micromirror device that rapidly switches binary coding patterns during camera exposure, effectively compressing temporal information into a single snapshot. ”Compressed data are then processed by an EfficientSCI neural network that reconstructs high-quality video sequences through sophisticated spatiotemporal modeling,” Yuan added.
Notably, the system was able capture the Leidenfrost effect, a phenomenon where droplets levitate on vapor layers above superheated surfaces.
“The system clearly resolved droplet dynamics including impact, deformation, and bouncing behavior with sub-millisecond precision,” said first author Yunfeng Song. “Quantitative analysis showed excellent agreement with commercial high-speed cameras.”
“Being able to resolve these rapid thermofluidic phenomena effectively is useful for studying transient physical processes,” Professor Yuan added. “This technology matches the performance of costly high-speed cameras, but with a much lower cost and reduced data storage requirements—offering a more affordable choice for advanced imaging.”
The research team validated their system through rigorous spatial resolution tests, achieving 5μm resolution using standard USAF 1951 targets. The system's 20:1 compression ratio enables reconstruction of twenty high-speed frames from a single 263 kB measurement, producing 5.26 MB of video data with minimal quality loss.
###
Contact the author: Xin Yuan, School of Engineering, Westlake University, Hangzhou 310030, China, xyuan@westlake.edu.cn
The publisher KeAi was established by Elsevier and China Science Publishing & Media Ltd to unfold quality research globally. In 2013, our focus shifted to open access publishing. We now proudly publish more than 200 world-class, open access, English language journals, spanning all scientific disciplines. Many of these are titles we publish in partnership with prestigious societies and academic institutions, such as the National Natural Science Foundation of China (NSFC).
Article Title
Deep learning based video snapshot compressive imaging of Leidenfrost droplets
COI Statement
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.