Light scattering control in transmission and reflection with neural networks

Area of Science:

• Optics
• Machine Learning
• Biomedical Imaging

Background:

• Light scattering significantly hinders controlled light delivery in various applications like biomedical imaging and optical communications.
• Shaping the light wavefront before it enters a scattering medium is crucial for overcoming these limitations.

Purpose of the Study:

• To develop a machine-learning approach for controlling light wavefronts to mitigate scattering effects.
• To demonstrate the capability of neural networks in establishing relationships between transmitted and reflected light patterns for light control.

Main Methods:

• Training neural networks (NNs) using pairs of binary intensity patterns and corresponding intensity measurements.
• Utilizing NNs to determine wavefront corrections for shaping light after it passes through a scattering medium.
• Investigating the functional relationship between transmitted and reflected speckle patterns.

Main Results:

• Successfully trained NNs to correct wavefronts and shape light beams post-scattering.
• Demonstrated that NNs can identify a reliable relationship between transmitted and reflected light speckle patterns.
• Achieved focusing and scanning of light in transmission through opaque media by utilizing reflected light signals.

Conclusions:

• Neural networks offer a versatile and efficient solution for light shaping and scattering correction.
• The study validates the feasibility of controlling light transmission through opaque media by analyzing reflected light.
• This approach has significant implications for applications requiring precise light delivery in scattering environments.