Related Experiment Video
Updated: Dec 9, 2025
09:04
Lens-free Video Microscopy for the Dynamic and Quantitative Analysis of Adherent Cell Culture
Published on: February 23, 2018
9.8K
Direct and accurate phase unwrapping with deep neural network.
Applied Optics
|September 9, 2020
Summary
A new deep neural network (DNN), VUR-Net, achieves accurate phase unwrapping directly, even with noisy or degraded data. This novel method outperforms existing techniques and introduces new quality evaluation indices.
Area of Science:
- Computer Vision
- Machine Learning
- Signal Processing
Background:
- Phase unwrapping is crucial for many imaging applications.
- Existing methods often require preprocessing or struggle with degraded data.
Purpose of the Study:
- To propose VUR-Net, a novel deep neural network for direct and accurate phase unwrapping.
- To evaluate VUR-Net's performance against state-of-the-art methods.
Main Methods:
- Developed VUR-Net, a deep neural network (DNN) utilizing numerous filters and alternating residual blocks.
- Compared VUR-Net with two other leading phase unwrapping DNNs.
- Introduced two new indices for evaluating phase unwrapping quality.
Main Results:
- VUR-Net achieved precise phase unwrapping without preprocessing or postprocessing.
- The network demonstrated robustness against noise, undersampling, and deformation.
- VUR-Net significantly outperformed existing DNNs in accuracy and robustness.
Conclusions:
- VUR-Net offers a superior solution for direct and accurate phase unwrapping.
- The proposed evaluation indices are effective for assessing phase unwrapping quality.
