Diffuse light field imaging through thick scattering media.

Area of Science:

• Optics and Photonics
• Image Reconstruction
• Scattering Media Imaging

Background:

• Imaging through scattering media is challenging due to low signal-to-noise ratios (SNRs).
• Ballistic photons, crucial for conventional imaging, become negligible in thick scattering media.
• Existing methods struggle with extremely low SNR conditions.

Purpose of the Study:

• To develop a novel imaging model for reconstructing clear images from scattering media at ultra-low SNRs.
• To retrieve forward-scattered photons as usable signals for image reconstruction.
• To extend light field imaging capabilities into diffuse environments.

Main Methods:

• Proposed a diffuse light field imaging model (DLIM).
• Constructed a diffuse source in an angular accumulated light field image by synthesizing radiance from multiple views.
• Solved for the clear image analytically using a diffusion equation and deconvoluting a diffuse Green function.
• Introduced three-plane parametrization for mathematical demonstration.

Main Results:

• DLIM successfully reconstructs target objects from scattering light field images even when they appear as random noise.
• The model achieves significantly higher performance compared to state-of-the-art methods.
• Outperformed peplography by 1.70 dB/4.76 dB PSNR and 0.167/0.172 SSIM for passive-luminous/self-luminous targets, respectively.

Conclusions:

• DLIM is the first physically-aware scattering light field imaging model.
• It effectively extends conventional light field imaging from free space to diffuse media.
• The model demonstrates robust performance in extremely low SNR scattering environments.