High speed single pixel imaging using a microLED-on-CMOS light projector

Area of Science:

• Optics and Photonics
• Computational Imaging
• Materials Science

Background:

• Single pixel imaging (SPI) systems typically face trade-offs between speed and flexibility.
• Existing SPI methods often require complex optics or limited hardware pattern sets.
• High frame-rate operation is crucial for dynamic scene capture.

Purpose of the Study:

• To demonstrate a flexible, high frame-rate single pixel imaging system.
• To leverage a novel microLED projector for enhanced imaging performance.
• To explore the potential of deep learning-generated patterns for SPI.

Main Methods:

• Utilized a newly developed microLED light projector with individually addressable pixels.
• Implemented Hadamard and Noiselet patterns for image reconstruction.
• Generated custom pattern sets using deep learning tools tailored to the microLED projector.
• Achieved image reconstruction from single-pixel measurements.

Main Results:

• Demonstrated single pixel imaging at pattern frame-rates approaching 400 kfps.
• Successfully generated 128x128 pixel images at 7.3 fps using standard pattern sets.
• Achieved nearly 800 fps imaging using deep learning-optimized patterns.
• The microLED projector enabled faster pattern modulation than traditional devices.

Conclusions:

• The microLED-based single pixel imaging system offers unprecedented speed and flexibility.
• Deep learning-generated patterns significantly enhance SPI performance at high frame rates.
• This technology paves the way for advanced dynamic imaging applications.