Noise-Resilient Imaging through Coherence Filtering

TL;DR

This paper introduces a coherence-based imaging method that effectively filters noise by exploiting differences in temporal coherence, outperforming traditional and quantum techniques especially under challenging noise conditions.

Contribution

The authors propose a novel interferometric coherence filtering approach that enhances noise resilience in imaging, applicable to various fields and overcoming limitations of existing methods.

Findings

Successfully recovered QR codes and grayscale wheels obscured by intense noise

Outperformed spectral filtering in regimes with substantial spectral overlap

Demonstrated robustness in imaging under high noise levels and low-light conditions

Abstract

Noise is a significant challenge in imaging. Conventional intensity-based techniques mitigate noise through various filtering methods, but they often require prior knowledge of noise characteristics and struggle, especially under low-light conditions and with spatially structured noise. Quantum distillation provides enhanced noise rejection; however, its applicability is limited as it requires specialised illumination and substantial modifications to existing imaging setups. In this article, we introduce a coherence-based image distillation approach that separates object from noise by leveraging the difference in their temporal coherence properties. We implement this through our interferometric protocol, which enables imaging based on spatial coherence while simultaneously filtering out noise via temporal coherence. This overcomes the limitations of both intensity-based and quantum distillation methods. We experimentally demonstrate noise resilience by successfully recovering feature-rich objects, such as QR codes and grayscale wheels, obscured by spatially uniform and structured noise 20 times as intense as the object. We further show that our method remains effective for fields with substantial spectral overlap, outperforming spectral filtering in regimes where the latter provides little noise suppression. This approach provides a robust framework for noise-resilient imaging with applications in optical communication, fluorescence microscopy, and biological imaging at both high and low light levels.