High-Resolution and Ultralow-Power Nonlinear Image Processing with Passive High-Quality Factor Metasurfaces

Area of Science:

• Photonics and Nanotechnology
• Artificial Intelligence
• Image Processing

Background:

• Artificial intelligence (AI) image processing is computationally intensive.
• Nanostructured metasurfaces offer potential for energy-efficient, direct image data processing.
• A key challenge is achieving nonlinear thresholding functions at low intensities without external power.

Purpose of the Study:

• To present a passive, all-optical method for nonlinear image processing.
• To demonstrate nonlinear thresholding functions using silicon nanoantenna arrays.
• To enable low-intensity image processing for energy-efficient AI applications.

Main Methods:

• Utilizing silicon nanoantenna arrays for nonlinear image processing.
• Leveraging opto-thermal nonlinearity via high-Q guided mode resonance.
• Experimentally demonstrating an intensity thresholding filter for one-dimensional images.

Main Results:

• Achieved an experimental threshold as low as 0.1 mW/μm².
• Demonstrated a spatial resolution of 1.85 μm.
• Simulations indicate potential for further reduced thresholds with maintained spatial selectivity.

Conclusions:

• The developed method provides a passive, all-optical approach for nonlinear image processing.
• Silicon nanoantenna arrays enable low-intensity thresholding without external power.
• Analog, pixel-wise activation filters could revolutionize image sensing technologies.