Deep learning sheds new light on non-orthogonal optical multiplexing.

Area of Science:

• Optics and Photonics
• Machine Learning
• Image Processing

Background:

• Speckle imaging through multimode fibers is challenging due to signal degradation.
• Non-orthogonal encoding complicates image recovery in scattering media.

Purpose of the Study:

• To develop a deep neural network for reconstructing speckle images from non-orthogonal input channels.
• To explore novel optical multiplexing techniques over scattering media.

Main Methods:

• Implementation of a deep neural network architecture.
• Utilizing non-orthogonal input channel encoding.
• Image recovery through a multimode fiber.

Main Results:

• Successful recovery of speckle images.
• Demonstration of a functional deep learning approach for this task.

Conclusions:

• The proposed deep neural network effectively recovers speckle images.
• This approach offers new possibilities for non-orthogonal optical multiplexing in scattering environments.