[Paper Review] Better than a lens - A novel concept to break the SNR-limit, given by Fermat's principle

J. Becker, Ronny Förster+1 more|arXiv (Cornell University)|Nov 20, 2018

Advanced Fluorescence Microscopy Techniques14 references5 citations

TL;DR

This paper proposes a novel single-shot imaging technique that breaks the fundamental SNR limit imposed by Fermat's principle and photon noise distribution in Fourier space. By splitting the pupil into two sub-pupils, capturing sub-images with reduced noise variance, and recombining them via weighted averaging in Fourier space, the method enhances Fourier-SNR and increases effective resolution beyond conventional limits, validated experimentally using a phase-only SLM with orthogonal blazed gratings.

ABSTRACT

The signal in an incoherent imaging system gets transferred worse towards higher spatial frequencies. Since the photon noise is distributed equally in Fourier space, the signal-to-noise ratio (SNR) decreases, and therefore represents a fundamental limit in the field of imaging. Our work aims to go beyond this given SNR-limit, which is directly connected to circumvent Fermats principle (1662). The proposed technique only makes use of a single acquisition process. To our knowledge this hasnt been done before and might have far-reaching impact on a large number of applications ranging from bio-imaging to industrial inspection. Especially in the case of limited photon numbers, e.g. fluorescent emission or a minimum available detector exposure time (temporal resolution), our method represents a new way to improve on SNR, which has been missing so far. We propose a novel and fundamentally new concept, by splitting the pupil of an imaging system into two parts to simultaneously obtain sub-images via each of these sub-pupils. Since the noise variance is proportional to the number of detected photons, it has been reduced in each of the sub-images. Recombining them by weighted averaging in Fourier-space yields an enhancement in (Fourier-) SNR compared to conventional imaging, enabling us to break the aforementioned SNR-limit. This results in an increased effective resolution limit in real space, which is used as an indicator throughout this work. We show a proof-of-concept experiment, in which we split the pupil by displaying orthogonally oriented blazed phase gratings for the different sub-pupils on a phase-only spatial light modulator. Our experiment confirms the improvement in effective resolution and therefore shows that we have overcome the fundamental SNR-limit, by only manipulating the captured light distribution in the pupil plane.

Motivation & Objective

• To overcome the fundamental SNR limit in incoherent imaging systems, which arises from equal photon noise distribution in Fourier space.
• To address the challenge of limited photon counts in applications like fluorescence microscopy and low-light industrial inspection.
• To develop a single-acquisition method that enhances effective resolution without violating Fermat's principle.
• To demonstrate a fundamentally new approach to SNR improvement that does not rely on multiple exposures or complex computational reconstruction.

Proposed method

• The pupil of the imaging system is split into two sub-pupils using a phase-only spatial light modulator (SLM).
• Orthogonal blazed phase gratings are displayed on the SLM to direct light into two distinct sub-pupil regions.
• Two sub-images are captured simultaneously, each with reduced noise variance due to the lower photon count per sub-pupil.
• The sub-images are transformed into Fourier space, where they are recombined using a weighted average to enhance the Fourier-SNR.
• The recombination process leverages the statistical advantage of combining independent, low-noise measurements to improve overall signal fidelity.
• The method avoids violating Fermat's principle by manipulating the pupil plane distribution rather than altering the optical path length in a way that contradicts ray-optics.

Experimental results

Research questions

• RQ1 Can the fundamental SNR limit in incoherent imaging be overcome using a single acquisition and pupil-plane manipulation?
• RQ2 Does splitting the pupil into sub-pupils and recombining sub-images in Fourier space lead to a measurable improvement in effective resolution?
• RQ3 To what extent can noise variance reduction in sub-images enhance the Fourier-SNR compared to conventional imaging?
• RQ4 Is it possible to achieve SNR enhancement without increasing photon exposure or using multiple acquisitions?
• RQ5 Can this method be experimentally validated using a phase-only SLM with orthogonal blazed gratings?

Key findings

• The experimental setup successfully demonstrated improved effective resolution by manipulating the pupil plane with a phase-only SLM.
• The use of two sub-pupils reduced noise variance in each sub-image, enabling a higher Fourier-SNR upon recombination.
• The weighted averaging of sub-images in Fourier space led to a measurable enhancement in signal fidelity beyond conventional imaging limits.
• The method achieved SNR improvement without violating Fermat's principle, indicating a fundamental departure from traditional imaging constraints.
• The proof-of-concept experiment confirmed that the proposed technique breaks the conventional SNR limit in a single acquisition, validating the theoretical framework.