Related Experiment Video
Updated: Jan 4, 2026
11:34
High-resolution, High-speed, Three-dimensional Video Imaging with Digital Fringe Projection Techniques
Published on: December 3, 2013
16.0K
Label enhanced and patch based deep learning for phase retrieval from single frame fringe pattern in fringe
Optics Express
|November 6, 2019
Summary
This study introduces a novel deep learning method for fast and accurate phase retrieval using fringe patterns. The approach enhances data and uses patches for improved 3D measurement accuracy.
Area of Science:
- Optics and Photonics
- Computer Vision
- Machine Learning
Background:
- Phase retrieval is crucial for 3D measurement using fringe projection.
- Existing deep learning methods often require extensive training data.
- Limitations in speed and accuracy persist in current phase retrieval techniques.
Purpose of the Study:
- To develop a fast and accurate deep learning-based phase retrieval method.
- To demonstrate the efficacy of label enhancement and patch-based strategies in fringe projection.
- To enable precise 3D measurement with limited training data.
Main Methods:
- A label-enhanced, patch-based deep learning approach for phase retrieval.
- Utilizing a deep neural network (DNN) trained on enhanced fringe patterns.
- Cropping training data into small, overlapped patches to augment sample size.
Main Results:
- Achieved fast and accurate phase retrieval with a small training dataset.
- Demonstrated the effectiveness of label enhancement and patch strategies for the first time in fringe projection deep learning.
- Successfully applied the method to experimental fringe patterns for dynamic 3D measurement.
Conclusions:
- The proposed method offers a significant advancement in deep learning-based phase retrieval.
- Label enhancement and patch strategies are key to improving DNN performance in fringe projection.
- The approach is suitable for dynamic fringe projection 3D measurement applications.
