Imaging through scattering media using speckle pattern classification based support vector regression.

Area of Science:

• Optics and Photonics
• Biomedical Imaging
• Machine Learning

Background:

• Imaging through scattering media is crucial for biomedical applications.
• Scattering media degrade object images into unrecognizable speckle patterns.
• Existing methods for image reconstruction from speckle patterns have limitations.

Purpose of the Study:

• To demonstrate a single-shot imaging method for reconstructing images through scattering media.
• To address the limitations of existing methods by utilizing speckle pattern classification.
• To improve the fidelity of reconstructed images from scattering media.

Main Methods:

• A novel method based on classification and support vector regression (SVR) of measured speckle patterns.
• Demonstration of speckle pattern classification feasibility with presented formulas.
• Evaluation of imaging capabilities with and without speckle pattern classification.

Main Results:

• Speckle patterns can be effectively classified even when object images are unavailable.
• The proposed speckle pattern classification-based SVR method overcomes prior limitations.
• High-fidelity reconstruction of object images through scattering media was achieved experimentally.

Conclusions:

• The developed method enables high-fidelity single-shot imaging through scattering media.
• Speckle pattern classification is a viable approach for image reconstruction.
• The technique holds promise for diverse sensing applications.