Physical twinning for joint encoding-decoding optimization in computational optics: a review

Area of Science:

• Computational optics
• Artificial intelligence in optics
• Deep learning for optical imaging

Background:

• Traditional optical systems face limitations in sensing dimension, light throughput, and resolution.
• Computational optics integrates computation to overcome these limitations, enhancing imaging and sensing.
• Deep learning has significantly advanced computational optics, improving precision and efficiency.

Purpose of the Study:

• To explore optical modulation elements for digital twin models in joint encoding-decoding optimization.
• To address the challenge of reverse physical twinning from optimized parameters to practical modulation elements.
• To provide guidance for selecting appropriate modulation elements in computational optics.

Main Methods:

• Review of various optical modulation elements across spatial, phase, and spectral dimensions.
• Analysis of digital twin models for joint encoding-decoding optimization.
• Examination of the discrepancies between optimized encoding parameters and physical modulation elements.

Main Results:

• Identification of challenges in reverse physical twinning due to gaps in bit depth, numerical range, and stability.
• Exploration of diverse optical modulation elements within the digital twin framework.
• Analysis of trade-offs between precision, speed, and robustness for different modulation elements.

Conclusions:

• The digital twin model offers enhanced performance for computational optics.
• Guidance is provided for selecting modulation elements based on specific imaging and sensing task requirements.
• This review aims to facilitate the development of next-generation computational optics by addressing twinning challenges.