Central wavelength estimation in spectral imaging behind a diffuser via deep learning.

Area of Science:

• Optics and Photonics
• Computer Vision
• Machine Learning

Background:

• Multispectral imaging through scattering media is crucial for sensing applications.
• Speckle images contain encoded spatial and spectral information, enabling object structure and spectrum estimation.
• Current methods face challenges in reconstructing images and estimating spectral properties through scattering media.

Purpose of the Study:

• To develop a deep learning-based strategy for estimating the central wavelength of illumination light from speckle images.
• To enable multispectral imaging through scattering media using a single monochrome camera.
• To reconstruct images and estimate spectral information in a single shot.

Main Methods:

• Utilizing deep learning to analyze speckle images captured with a monochrome camera.
• Training the model on speckle images acquired with narrowband light at different central wavelengths.
• Developing a strategy to extend the spectral region for image reconstruction and wavelength estimation.

Main Results:

• Accurate estimation of central wavelengths in 1 nm steps for objects with a 100 nm spectral variation.
• Successful image reconstruction and central wavelength estimation in a single shot.
• Demonstration of deep learning's capability to decode spectral information from speckle patterns.

Conclusions:

• The proposed deep learning method effectively estimates central wavelengths from speckle images, overcoming scattering media challenges.
• This technique facilitates single-shot multispectral imaging through scattering media using monochrome cameras.
• The approach holds significant potential for advancing sensing technologies in various fields.