01 Overview
The research team led by Professor Zeng Heping and Researcher Huang Kun at East China Normal University has made significant progress in high-speed mid-infrared imaging. They proposed a novel approach based on nonlinear spatial multiplexing, overcoming the longstanding trade-off between frame rate and pixel count in conventional architectures. This method achieves a tenfold improvement in imaging speed while maintaining megapixel-level spatial resolution, enabling high-fidelity mid-infrared video imaging at 10 kHz. The work supports real-time observation in combustion diagnostics, explosion monitoring, and thermal transport analysis. The results, titled “High-Speed Mid-Infrared Imaging via Nonlinear Multiplexed Detection,” have been published in Laser & Photonics Reviews. East China Normal University is the leading affiliation. Doctoral student Qin Ruiyang is the first author, with Researcher Huang Kun and Professor Zeng Heping serving as corresponding authors.
Fig 1: Laser & Photonics Reviewspublishes the latest results of the research team of East China Normal University
02 Background
High-speed mid-infrared imaging is a key enabling technology for capturing dynamic phenomena in target tracking, transient physics, and complex system analysis. However, existing large-format mid-infrared detectors face critical limitations in pixel scalability, readout speed, and sensitivity, hindering the ability to simultaneously achieve high frame rates and high spatial resolution. This severely restricts real-time imaging in rapidly evolving or non-repeatable scenarios.
Current high-speed imaging techniques are primarily developed for the visible and near-infrared regions. In the mid-infrared range, two fundamental challenges persist: (1) the lack of detectors offering both high sensitivity and temporal resolution, and (2) the immaturity of high-resolution dynamic spatial modulation methods at longer wavelengths, where diffraction limits precision and stability. Frequency upconversion imaging has recently emerged as a promising solution, enabling indirect sensing of mid-infrared fields by converting them to the visible or near-infrared using nonlinear optics. However, the achievable frame rate remains constrained by the intrinsic readout speed of the camera, with improvements often at the cost of reduced pixel count. Thus, achieving both high temporal and spatial resolution in mid-infrared imaging remains a major challenge.
Fig 2: Conceptual illustration of the high-speed mid-infrared imaging via nonlinear multiplexed detection
03 Technical Innovation
To address this, the team introduced a nonlinear spatial multiplexed imaging scheme (Fig. 3), combining high frame rate, wide field of view, and high sensitivity. The method enables mid-infrared imaging at 10 kHz frame rates with megapixel resolution. Specifically, spatially structured pump patterns with distinct spatial frequencies and orientations were generated using a high-speed programmable digital micromirror device (DMD). During the nonlinear sum-frequency generation process, both spatial encoding and frequency upconversion were simultaneously achieved, enabling all-optical control of the mid-infrared field.
Leveraging the mutual orthogonality of the pump patterns in spatial frequency and the natural low-frequency sparsity of real-world scenes, image frames at different time points were projected into non-overlapping spectral bands in the Fourier domain. This allowed time-domain information to be multiplexed spatially and spectrally in a single exposure. In the reconstruction stage, a 2D Fourier transform separated the frequency components, followed by bandpass filtering and inverse Fourier transform to recover the transient frame sequence.
Fig 3: Experimental setup
In the experiment (Fig. 4), the DMD operated at 10 kHz, and the camera captured images at 1 kHz with 1 ms exposure time. This enabled encoding of ten frames into each camera exposure. Compared to conventional imaging without modulation, the method effectively mapped temporal dynamics into spatial frequency shifts, mitigating motion blur and preserving structural features. Fourier analysis revealed ten clearly separated first-order harmonics, confirming excellent spectral isolation between channels. After demodulation, a 10 kHz equivalent frame sequence was successfully reconstructed, boosting frame rate by an order of magnitude without compromising resolution. Further improvements—such as narrowing filter bandwidth or increasing system numerical aperture—could expand the number of channels, paving the way for even higher frame rates.
Fig 4: Experimental results
04 Summary and Outlook
By integrating nonlinear frequency conversion with spatial pump encoding, this study demonstrates—for the first time—10 kHz mid-infrared imaging at megapixel resolution, breaking through the traditional trade-off between speed and resolution. The system also supports continuous streaming acquisition, enabling real-time recording of long-duration, high-speed transient events such as combustion or explosions.
Looking ahead, the method can benefit from faster spatial modulators, compressed sensing strategies, and deep learning-based reconstruction to further enhance frame rate. Owing to its wavelength compatibility, the technique is also extendable to long-wave infrared and terahertz bands, addressing urgent needs for high-speed, high-resolution imaging across a broader spectral range.
The team has previously developed several pioneering techniques in mid-infrared nonlinear imaging and sensing, including nonlinear Fourier layered imaging [Optica 11, 1716 (2024)], mid-infrared hyperspectral video imaging [Nat. Commun. 15, 1811 (2024)], and single-photon single-pixel imaging [Nat. Commun. 14, 1073 (2023)]. These efforts have been supported by the Ministry of Science and Technology, the National Natural Science Foundation of China, Shanghai Municipal Science and Technology Commission, Chongqing Science and Technology Bureau, and East China Normal University.
Link to paper: High-Speed Mid-Infrared Imaging via Nonlinear Multiplexed Detection
