EDoF-ToF: extended depth of field time-of-flight imaging

Area of Science:

• Computer Vision
• Computational Imaging
• Sensor Technology

Background:

• Conventional continuous-wave amplitude-modulated time-of-flight (CWAM ToF) cameras face a trade-off between light throughput and depth of field (DoF).
• Larger apertures increase light collection and signal-to-noise ratio but drastically reduce DoF, limiting practical applications.
• Both high light throughput and wide DoF are crucial for robust CWAM ToF performance.

Purpose of the Study:

• To develop an algorithmic method, EDoF-ToF, to extend the DoF of large-aperture CWAM ToF cameras.
• To enable CWAM ToF systems to achieve an all-in-focus measurement by overcoming the limitations of narrow focal regions.
• To provide a software-only solution for extended DoF in ToF imaging.

Main Methods:

• Proposed EDoF-ToF, a neural network-based approach to deblur objects outside the lens's focal region.
• Developed a novel training data simulator for large-aperture ToF cameras, modeling depth-dependent nonlinear blurs and partial occlusions.
• Derived a nonlinear blur model specific to ToF depth maps, distinct from linear models in conventional deblurring.

Main Results:

• Experimentally demonstrated a 3.6x increase in the DoF of a conventional ToF system using EDoF-ToF.
• Achieved the DoF equivalent to a smaller aperture system that collects 22.1x less light.
• Validated the effectiveness of the nonlinear blur model and the training data simulator.

Conclusions:

• EDoF-ToF successfully extends the DoF of large-aperture CWAM ToF cameras.
• The method allows ToF cameras to benefit from both high light throughput and a wide DoF simultaneously.
• This software solution eliminates the need for additional hardware typically required for extended DoF photography.