Performance Comparison of Ultra-Miniature Diffraction Gratings with Lenses and Zone Plates

Leonidas Spinoulas, Oliver Cossairt, Aggelos K. Katsaggelos, P. Gill, David G. Stork2015 · DOI

TL;DR

This paper compares ultra-miniature diffraction gratings with traditional lenses and zone plates for computational imaging. The novel gratings offer a surprisingly uniform reconstruction quality across a wide field of view and varying wavelengths, potentially outperforming traditional optics in compact imaging systems despite their non-focusing nature.

What Problem Does This Paper Address?

Traditional imaging systems rely on lenses or zone plates to focus light, creating a sharp Point Spread Function (PSF) to form an image. However, miniaturizing these optical elements for applications like endoscopy or surveillance presents significant fabrication challenges, leading to deviations from ideal performance. This paper investigates an alternative: ultra-miniature diffraction gratings that, instead of focusing, impose phase anti-symmetry on the incoming wavefront. These gratings produce a wide PSF but retain high-frequency information and are designed to be relatively insensitive to manufacturing imperfections and wavelength variations. The core problem addressed is understanding if these non-focusing gratings can offer comparable or superior imaging performance to traditional optics in a miniaturized, computationally reconstructed imaging system, especially considering their potential robustness to fabrication defects.

Methods

The authors model an imaging system consisting of a sensor, a propagation medium with a specific refractive index, and an optical element (spiral grating, Fresnel zone plate, or singlet lens). The system is designed for far-field imaging, assuming incoming light forms plane waves of different angles. Fraunhofer diffraction is used for propagation through the grating, and Fresnel diffraction models propagation within the medium to the sensor.

The core of the comparison lies in a computational reconstruction approach. The imaging process is modeled as a linear operation: $y = Ax + n$, where $y$ is the sensor measurement, $A$ is the diffraction response matrix (capturing the system's behavior for different incoming angles), $x$ is the unknown image, and $n$ is noise. Tikhonov regularization is employed to solve for $x$, with a regularization parameter $\gamma$ tuned to the noise level. The authors also account for light falloff proportional to $(\cos\theta)^4$, where $\theta$ is the incoming light angle, and correct for this effect during reconstruction. Performance is evaluated using Peak Signal-to-Noise Ratio (PSNR) across different wavelengths (400, 500, and 600 nm) and increasing noise levels.

Datasets

Not discussed in the paper. The experiments utilize simulated data based on the described imaging system model.

Key Insights and Ideas

  • Uniformity is Key: The spiral gratings demonstrate remarkably uniform reconstruction quality across a wide field of view, unlike lenses and zone plates which suffer from significant blurring or corruption at the edges.
  • Wavelength Independence: The performance of the spiral gratings remains consistent across different wavelengths (400, 500, and 600 nm), a significant advantage over traditional optics which are typically optimized for a specific wavelength.
  • Robustness to Imperfections: The paper suggests that the spiral grating's design, which is less sensitive to manufacturing defects and depth variations compared to lenses and zone plates, could be a major benefit in ultra-miniature systems where fabrication is challenging.
  • Computational Imaging Advantage: By embracing a wide PSF and relying on computational reconstruction, these gratings can achieve good performance even with non-ideal optical elements, potentially offering a more practical solution for miniaturization.

Key Results

  • The spiral gratings consistently achieve comparable or better PSNR values than Fresnel zone plates and singlet lenses across various wavelengths and noise levels.
  • Qualitative results (Figure 3) show that while lenses and zone plates produce blurred edges, the spiral grating provides a more uniform reconstruction, effectively resolving high frequencies across the entire field of view, albeit with slightly lower quality at very small FOVs.
  • The wavelength independence of the spiral grating's performance is visually evident in the reconstructed images, which show less degradation compared to the zone plate and lens when the wavelength deviates from the optimization target.

Limitations and Future Work

Limitations: * The paper acknowledges that the tiny scale of the imagers makes fabrication challenging, and defects can lead to deviations from optimal PSFs. * The current reconstruction technique might exhibit ringing artifacts at high angles, which could be mitigated by more sophisticated, edge-preserving algorithms.

Future Work: * Exploring more advanced reconstruction techniques to further improve image quality and reduce artifacts, especially at the edges of the field of view. * Investigating the practical fabrication challenges and their impact on the performance of these ultra-miniature gratings. * Further optimizing the grating design for specific applications beyond general imaging, such as QR-code reading or face detection.

Summary updated

Abstract

In this work, we compare the performance of previously proposed ultra-miniature diffraction gratings with ideal lenses and zone plates of similar structural characteristics. The analysis aims at understanding the differences of designs utilizing non-focusing gratings and the potential benefits of their use in computational imaging systems.
BibTeX
@misc{spinoulas2015performance,
  author = {Leonidas Spinoulas and Oliver Cossairt and Aggelos K. Katsaggelos and P. Gill and David G. Stork},
  title = {{Performance Comparison of Ultra-Miniature Diffraction Gratings with Lenses and Zone Plates}},
  year = {2015},
  doi = {10.1364/cosi.2015.cm3e.1},
}
Figure