All-Optical Diffractive Operators for Rapid, Computer-Free Morphological Transformations

Area of Science:

• Optics and Photonics
• Computer Vision
• Deep Learning

Background:

• Morphological transformations are crucial for image processing but computationally intensive.
• Existing digital methods require significant memory and processing power, especially for large datasets.

Purpose of the Study:

• To develop a fast, highly parallel, and computer-free method for morphological transformations using diffractive computing.
• To demonstrate the flexibility and scalability of this all-optical approach for various image processing applications.

Main Methods:

• Utilizing cascaded diffractive surfaces designed via a deep learning-based optimization process.
• Implementing free-space diffractive devices to process optical wavefronts directly for dilation and erosion.
• Employing a reflection configuration with a phase-only spatial light modulator (SLM).

Main Results:

• Successfully performed morphological transformations (dilation and erosion) on amplitude- and phase-encoded images.
• Demonstrated computer-free, all-optical processing with high parallelism and scalability.
• Showcased image denoising and flexible tuning of transformation kernels by adjusting training datasets.

Conclusions:

• Diffractive computing offers an efficient and scalable alternative to digital methods for morphological transformations.
• The developed all-optical processor enables computer-free, real-time image processing with tunable functionalities.
• This approach has significant potential for applications in bioimaging, surveillance, and environmental monitoring.