Color-resolved light field shaping via diffractive-electronic U-shape network with wavelength-aware virtual branching

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Computational Imaging

Background:

• Dynamic spatial distortions challenge stable optical field shaping, especially for color patterns.
• Current techniques, including diffractive optical neural networks, are limited to monochromatic targets and struggle with temporal variations.

Purpose of the Study:

• To develop a novel diffractive-electronic hybrid neural network for real-time, color light field shaping.
• To address limitations in spectral generalization and temporal stability in existing beam shaping methods.

Main Methods:

• Introduced a wavelength-aware virtual branching (WAVB) mechanism for spectral generalization.
• Integrated a spectrally conditioned U-shape network with frequency-selective skip connections (FSSC).
• Developed an all-optical-driven optical flow prediction module for temporal coherence.

Main Results:

• Achieved real-time operation at 50Hz for color light field shaping.
• Demonstrated robust, frame-stable shaping across diverse spatial and temporal distortion scenarios.
• The hybrid network successfully adapted to wavelength-specific shaping and preserved inter-channel dependencies.

Conclusions:

• The proposed system offers a scalable framework for intelligent, adaptive optical imaging.
• Enables robust, real-time color light field shaping, overcoming limitations of previous methods.
• Presents promising applications in dynamic holography, laser displays, and computational optical imaging.