Related Experiment Video
Updated: Aug 12, 2026
14:25
Determining 3D Flow Fields via Multi-camera Light Field Imaging
Published on: March 6, 2013
High Frame Rate iToF-Flow-Based Depth Imaging Enabled by Local Linear Transfer
IEEE Transactions on Pattern Analysis and Machine Intelligence
|August 10, 2026
Summary
This study introduces the generalized iToF-flow (GiF) model to improve 3D depth perception in dynamic scenes using indirect Time-of-Flight (iToF) technology. The GiF model significantly enhances depth reconstruction accuracy and reduces computation time for iToF systems.
Area of Science:
- Computer Vision
- 3D Perception Technologies
- Computational Imaging
Background:
- Indirect Time-of-Flight (iToF) is a cost-effective 3D perception technology.
- iToF performance degrades in dynamic scenes due to multiple measurements.
- Local Linear Transfer (LLT) characteristics and iToF imaging physics are key considerations.
Purpose of the Study:
- To develop a novel model addressing iToF limitations in dynamic environments.
- To enhance depth extraction accuracy and frame rates for iToF systems.
- To account for variations from different measurement modes and 3D motion.
Main Methods:
- Proposed the generalized iToF-flow (GiF) model with cross-mode and uni-mode flow.
- Designed the generalized iToF-flow-based depth extraction network (GiFDEN).
- Integrated an LLT-based cross-mode transfer module (ILCTM) and a uni-mode photometric compensation module (UPCM).
Main Results:
- GiFDEN accurately estimates four-phase measurements at each time step.
- Achieved high-frame-rate and accurate depth retrieval.
- Reduced computation time by 87% and depth reconstruction error by 55% compared to SOTA methods.
Conclusions:
- The proposed GiF model and GiFDEN network effectively improve iToF performance in dynamic scenes.
- The method demonstrates significant advancements in speed and accuracy for 3D depth extraction.
- Validated through extensive experiments on simulated and real-world data.
