Dual-domain mean-reverting diffusion model-enhanced temporal compressive coherent diffraction imaging

Area of Science:

• Optics and Photonics
• Image Processing
• Computational Imaging

Background:

• Temporal compressive coherent diffraction imaging (TCDI) is a lensless technique for capturing dynamic micro-scale objects.
• Image reconstruction quality in TCDI is often limited by the loss of frequency domain information.
• Existing methods struggle to fully recover lost data, impacting spatial resolution and signal-to-noise ratio.

Purpose of the Study:

• To introduce a novel method, dual-domain mean-reverting diffusion model-enhanced temporal compressive coherent diffraction imaging (DMDTC), for improved TCDI reconstruction.
• To leverage deep learning, specifically mean-reverting diffusion models, for data recovery and image enhancement in the frequency and spatial domains.
• To overcome the limitations of conventional TCDI by enhancing image quality, particularly structural similarity and peak signal-to-noise ratio.

Main Methods:

• Development of a dual-domain mean-reverting diffusion model (DMDTC) integrating frequency and spatial domain processing.
• Application of a frequency domain mean-reverting diffusion model to recover missing spectral information.
• Utilization of a hybrid input-output algorithm for spatial domain image reconstruction, followed by spatial domain mean-reverting diffusion model for denoising and restoration.

Main Results:

• DMDTC significantly enhances the quality of reconstructed images in temporal compressive coherent diffraction imaging.
• Reconstructed images using DMDTC demonstrate superior structural similarity and peak signal-to-noise ratio compared to conventional methods.
• The method successfully enables high temporal frame rates and high spatial resolution imaging.

Conclusions:

• DMDTC offers a substantial advancement in lensless imaging, particularly for dynamic micro-scale object capture.
• The integration of dual-domain diffusion models effectively addresses information loss in TCDI.
• This technique provides a robust solution for achieving high-fidelity imaging with both high temporal and spatial resolution.