Related Experiment Video
Updated: Sep 11, 2025
11:34
High-resolution, High-speed, Three-dimensional Video Imaging with Digital Fringe Projection Techniques
Published on: December 3, 2013
15.7K
Dynamic quantitative phase imaging using deep spatial-temporal prior
Optics Express
|August 13, 2025
Summary
This study introduces a faster deep learning method for quantitative phase imaging (QPI) using spatial-temporal prior (STeP) and a physics-enhanced neural network (PhysenNet). It significantly reduces computational costs and training time for dynamic object imaging.
Area of Science:
- Optics and Photonics
- Biomedical Imaging
- Artificial Intelligence
Background:
- Non-interferometric deep learning-based quantitative phase imaging (QPI) offers label-free optical path length delay measurements.
- Integrating deep learning with physical knowledge enhances QPI precision but faces lengthy optimization challenges, limiting multi-frame applications.
Purpose of the Study:
- To develop an efficient method for QPI of dynamic objects using physics-enhanced neural networks.
- To reduce computational costs and training time for multi-frame QPI tasks.
Main Methods:
- Leveraged spatial-temporal prior (STeP) from video sequences.
- Incorporated lightweight convolutional operations into a physics-enhanced neural network (PhysenNet).
- Applied the method to QPI of dynamic objects without additional measurements.
Main Results:
- Achieved accurate reconstructions of dynamic phase distributions.
- Reduced computational costs and training time by over 90%.
- Demonstrated effectiveness even under low signal-to-noise ratio conditions.
Conclusions:
- The STeP-enhanced PhysenNet offers a significant advancement in efficient QPI for dynamic objects.
- This method overcomes the limitations of lengthy optimization processes in previous deep learning QPI approaches.
- Paves the way for practical, efficient multi-frame inverse imaging solutions.
