Related Experiment Video
Updated: Sep 11, 2025
06:25
Time Multiplexing Super Resolving Technique for Imaging from a Moving Platform
Published on: February 12, 2014
8.5K
Efficient super-resolution of phase images encoded with random phase mask by machine learning techniques.
Applied Optics
|August 12, 2025
Summary
This study introduces a super-resolution method using machine learning to enhance phase image accuracy in optical measurements. The approach improves complex amplitude information encoding and decoding beyond device resolution limits.
Area of Science:
- Optics and Photonics
- Computational Imaging
- Machine Learning Applications
Background:
- Random phase masks (RPMs) encode spatial complex amplitude information into phase for efficient optical processing.
- Limitations in spatial light modulator (SLM) and charge-coupled device (CCD) resolution hinder encoding/decoding accuracy.
Purpose of the Study:
- To propose a super-resolution method for phase images to overcome optical device resolution limitations.
- To improve the accuracy of encoding and decoding spatial complex amplitude distribution.
Main Methods:
- Utilized convolutional neural networks (CNNs) and vision transformers (ViTs) for super-resolution processing of phase images.
- Applied the method to virtual phase conjugation based optical tomography (VPC-OT) for microscopic target measurement.
Main Results:
- Achieved super-resolution for complex amplitude information beyond SLM and CCD resolution limits.
- Demonstrated peak signal-to-noise ratio (PSNR) improvements of 2.37 dB (CNN) and 1.86 dB (ViT).
- Observed 6-13 dB measurement accuracy improvement and noise reduction in VPC-OT simulations.
Conclusions:
- The proposed machine learning-based super-resolution method enhances phase image accuracy without additional optical systems or increased measurements.
- The method offers improved measurement accuracy with minimal computational cost.
- Investigated optimal machine learning model sizing based on input image size and loss progression.
