A Diffuse Light Field Imaging Model for Forward-Scattering Photon-Coded Signal Retrieval

TL;DR

This paper introduces a physically-aware diffuse light field imaging model (DLIM) that effectively retrieves forward-scattered photons in super-strong scattering environments, significantly improving low-SNR imaging performance.

Contribution

The paper presents the first physically-aware scattering light field model based on diffusion equations, extending light field imaging into diffuse space for enhanced photon retrieval.

Findings

DLIM reconstructs objects at extremely low SNRs.

The model accurately models radiance distribution in scattering environments.

Experimental results demonstrate superior performance over traditional methods.

Abstract

Scattering imaging is often hindered by extremely low signal-to-noise ratios (SNRs) due to the prevalence of scattering noise. Light field imaging has been shown to be effective in suppressing noise and collect more ballistic photons as signals. However, to overcome the SNR limit in super-strong scattering environments, even with light field framework, only rare ballistic signals are insufficient. Inspired by radiative transfer theory, we propose a diffuse light field imaging model (DLIM) that leverages light field imaging to retrieve forward-scattered photons as signals to overcome the challenges of low-SNR imaging caused by super-strong scattering environments. This model aims to recover the ballistic photon signal as a source term from forward-scattered photons based on diffusion equations. The DLIM consists of two main processes: radiance modeling and diffusion light-field approximation. Radiate modeling analyzes the radiance distribution in scattering light field images using a proposed three-plane parameterization, which solves a 4-D radiate kernel describing the impulse function of scattering light field. Then, the scattering light field images synthesize a diffuse source satisfying the diffusion equation governing forward scattering photons, solved under Neumann boundary conditions in imaging space. This is the first physically-aware scattering light field imaging model, extending the conventional light field imaging framework from free space into diffuse space. The extensive experiments confirm that the DLIM can reconstruct the target objects even when scattering light field images are reduced as random noise at extremely low SNRs.