Restoration of motion-blurred numeral image using a complex-amplitude diffractive processor

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Image Processing

Background:

• Motion blur significantly degrades image quality in numeral recognition.
• Diffractive deep neural networks (D2NNs) offer a promising approach for optical information processing.

Purpose of the Study:

• To propose and evaluate a complex-amplitude diffractive processor for motion blur removal in numeral images.
• To compare the performance of complex-amplitude D2NNs against phase-only and amplitude-only counterparts.

Main Methods:

• Development of a complex-amplitude diffractive processor based on D2NNs.
• Precise control of optical field propagation for image restoration.
• Comparative analysis of different diffractive processor configurations (phase-only, amplitude-only, complex-amplitude).

Main Results:

• The complex-amplitude diffractive processor effectively removes motion blur and restores numeral images.
• Complex-amplitude networks significantly enhance processor performance and improve peak signal-to-noise ratio (PSNR).
• Reduced network layers and alleviated alignment difficulties were observed with complex-amplitude networks.

Conclusions:

• Complex-amplitude D2NNs offer superior performance for motion blur removal compared to other configurations.
• The proposed processor demonstrates fast processing speed and low power consumption.
• Potential applications include road monitoring, sports photography, satellite imaging, and medical diagnostics.