Event Cameras Meet SPADs for High-Speed, Low-Bandwidth Imaging

Area of Science:

• Computer Vision
• Sensor Technology
• Image Processing

Background:

• Conventional cameras struggle with a trade-off between low-light performance and high-speed imaging, leading to motion blur or noisy images.
• Single-photon avalanche diode (SPAD) sensors offer single-photon sensitivity and microsecond temporal resolution.
• Event cameras provide high temporal resolution (up to 1 MHz) with low bandwidth.

Purpose of the Study:

• To introduce a sensor fusion framework combining SPADs and event cameras.
• To improve low-light, high-speed image reconstruction.
• To reduce the bandwidth requirements of high-speed imaging systems.

Main Methods:

• Developed a sensor fusion framework integrating data from SPAD sensors and event cameras.
• Evaluated the framework using both synthetic and real sensor data.
• Compared reconstruction performance against conventional camera techniques.

Main Results:

• Achieved significant enhancements in image reconstruction quality (over 5 dB PSNR) for low-light scenes.
• Enabled high temporal resolution imaging (100 kHz) with improved fidelity.
• Demonstrated reduced bandwidth costs compared to using SPAD frames exclusively.

Conclusions:

• Event-SPAD sensor fusion effectively addresses the limitations of conventional cameras in challenging imaging conditions.
• This approach shows significant promise for real-world applications like robotics and medical imaging.
• The fusion framework offers a pathway to high-quality, high-speed, low-light imaging with efficient data handling.