Super-resolution depth imaging via processing of compact single-photon histogram parameters.

Area of Science:

• Computer Vision
• Sensor Technology
• Computational Imaging

Background:

• Time-of-flight (ToF) sensors are crucial for depth perception in consumer electronics.
• ToF data is often represented as photon arrival time histograms, enabling super-resolution techniques.
• Transferring full histogram data is challenging for compact systems due to high data volume.

Purpose of the Study:

• To investigate the feasibility of high-resolution depth imaging using minimal ToF sensor data.
• To develop a data-efficient neural network for enhancing ToF sensor spatial resolution.
• To demonstrate that extracted key parameters are sufficient for high-quality depth reconstruction.

Main Methods:

• Proposed a compact, data-efficient neural network architecture.
• Focused on extracting 3 key parameters (peak position, intensity, noise) per pixel.
• Enhanced spatial resolution from 4x4 to 32x32 pixels, a 48x data reduction compared to full histograms.

Main Results:

• Successfully reconstructed high-resolution depth images from reduced parameter data.
• Achieved performance comparable to methods using full histogram data.
• Demonstrated significant data reduction (48x) with minimal loss in depth imaging quality.

Conclusions:

• Key extracted parameters from ToF sensors are sufficient for high-resolution depth reconstruction.
• The proposed neural network offers an efficient solution for compact depth imaging systems.
• This approach enables high-quality, data-efficient depth sensing in consumer electronics.