Learning-based single-shot long-range synthetic aperture Fourier ptychographic imaging with a camera array.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning in Imaging

Background:

• Conventional macroscopic Fourier ptychographic imaging requires extensive data and high spectral overlap.
• Imaging diffuse reflective objects with high resolution and signal-to-noise ratio presents significant challenges.
• Existing methods are often limited in their ability to perform single-shot imaging, especially for dynamic scenes.

Purpose of the Study:

• To develop a new long-range synthetic aperture Fourier ptychographic imaging technique.
• To enable high-resolution, high signal-to-noise imaging of diffuse reflective objects in a single shot.
• To overcome limitations of conventional Fourier ptychographic imaging for macroscopic applications.

Main Methods:

• Development of learning-based single-shot synthetic aperture imaging (LSS-SAI).
• Utilizing a camera array to capture low-resolution intensity images across non-overlapping spectral regions in parallel.
• Reconstruction of super-resolved images using a physical model-based dual-regression deep neural network.

Main Results:

• LSS-SAI successfully reconstructs super-resolved images from parallel captured low-resolution spectral data.
• Experimental results show improvements in peak signal-to-noise ratio (PSNR) by 10.56 dB and structural similarity (SSIM) by 0.26 for rough reflective samples.
• Demonstrated single-shot ptychography of a dynamic scene at 30 fps, achieving camera-limited speed.

Conclusions:

• LSS-SAI offers a viable solution for single-shot, long-range, high-resolution imaging of diffuse reflective surfaces.
• The technique overcomes data acquisition and spectral overlap limitations of traditional methods.
• This work represents the first demonstration of macroscopic Fourier ptychography for single-shot imaging of dynamic events.