High-quality ghost imaging through highly complex scattering media with physics-enhanced untrained neural networks

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning Applications

Background:

• Optical imaging through scattering media is hindered by distorted illumination and detection paths.
• Ghost imaging (GI) performance degrades due to dynamic scaling factors in complex scattering environments.

Purpose of the Study:

• To develop a novel method for high-quality object reconstruction in optical imaging through complex scattering media.
• To overcome limitations of ghost imaging caused by simultaneous disturbances in illumination and detection paths.

Main Methods:

• Utilized a physics-enhanced untrained neural network (UNN) integrated with a ghost imaging (GI) model.
• Employed rotating ground glass diffusers and a turbidity-varying liquid chamber to create complex scattering.
• Recorded speckle patterns and single-pixel intensities, with UNN estimating dynamic scaling factors.

Main Results:

• Achieved robust and high-quality object reconstruction despite complex and dynamic scattering.
• Demonstrated effective compensation for beam distortions in both illumination and detection paths.
• Validated the capability of the physics-enhanced UNN to correct measurements for reliable reconstruction.

Conclusions:

• The proposed physics-enhanced UNN method enables robust optical imaging through simultaneously disturbed scattering media.
• This approach offers a promising solution for overcoming optical scattering challenges in diverse complex scenarios.
• Opens new avenues for advanced imaging techniques in turbid and dynamic environments.