Depth of field expansion method for integral imaging based on diffractive optical element and CNN.

Area of Science:

• Optics and Photonics
• Computer Vision
• Display Technology

Background:

• Lens-based display systems suffer from aberrations and limited depth of field (DoF), causing image blurring and distortion.
• Expanding DoF in displays often involves a trade-off between horizontal and axial resolution, hindering high-resolution 3D display capabilities.
• Existing methods struggle to achieve both extended DoF and high resolution simultaneously in 3D displays.

Purpose of the Study:

• To overcome the limitations of lens-based systems and enhance the DoF and resolution of reconstructed 3D images.
• To propose a novel method combining diffractive optical element (DOE) optimization and convolutional neural network (CNN) image pre-correction.
• To achieve depth-invariant point spread function (PSF) distribution and compensate for image quality degradation.

Main Methods:

• Diffractive optical element (DOE) phase distribution was optimized using the Adam algorithm, replacing conventional lenses.
• A convolutional neural network (CNN) was employed for pre-correcting original images to mitigate DOE-induced quality reduction.
• The combined method was implemented and tested within a practical integral imaging system.

Main Results:

• The proposed method successfully extended the DoF of the DOE to 400 mm.
• Achieved a high-resolution 3D display across multiple depth planes.
• Demonstrated depth-invariant and concentrated point spread function (PSF) distribution throughout the extended DoF range.

Conclusions:

• The integration of optimized DOE and CNN pre-correction effectively enhances DoF and resolution in 3D display systems.
• This approach overcomes the inherent trade-offs in conventional display technologies.
• Numerical simulations and optical experiments validated the effectiveness and practicality of the proposed method for advanced 3D displays.