Related Experiment Video
Updated: Jun 11, 2025
09:36
Characterization of Anisotropic Leaky Mode Modulators for Holovideo
Published on: March 19, 2016
7.9K
Frequency aware high-quality computer-generated holography via multilevel wavelet learning and channel attention
Optics Letters
|October 1, 2024
Summary
This study introduces a novel frequency-aware network for generating high-quality phase-only holograms (POHs) for holographic displays. The proposed multilevel wavelet-based channel attention network (MW-CANet) effectively addresses spectral bias in deep learning models.
Area of Science:
- Optics
- Computer Vision
- Artificial Intelligence
Background:
- Deep learning significantly advances computer-generated holography for real-time displays.
- Convolutional Neural Networks (CNNs) are common for phase-only hologram (POH) encoding but exhibit spectral bias, hindering high-frequency component learning.
Purpose of the Study:
- To propose a novel frequency-aware network to generate high-quality POHs.
- To address the spectral bias issue in deep learning models for holography.
Main Methods:
- Developed a multilevel wavelet-based channel attention network (MW-CANet).
- Employed multi-scale wavelet transformations to independently capture low- and high-frequency features.
- Integrated an attention mechanism to prioritize critical high-frequency components.
Main Results:
- The MW-CANet effectively captures both low- and high-frequency information.
- Enhanced representation of high-frequency components crucial for accurate phase inference.
- Simulations and optical experiments validated the method's effectiveness.
Conclusions:
- The proposed MW-CANet successfully generates high-quality POHs by mitigating spectral bias.
- This frequency-aware approach offers a significant improvement for deep learning-based holographic display technologies.
