Multi-depth hologram generation using stochastic gradient descent algorithm with complex loss function.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Computer Vision

Background:

• Stochastic gradient descent (SGD) is a key algorithm for phase-only hologram optimization, enabling high-quality holographic displays.
• Current SGD methods face significant computational challenges in multi-depth hologram generation, becoming impractical for complex 3D objects due to increasing optimization time with depth layers.

Purpose of the Study:

• To address the computational inefficiency of SGD in multi-depth hologram generation.
• To develop a method that significantly reduces optimization time while maintaining or improving holographic display quality for complex 3D objects.

Main Methods:

• Proposed a novel approach using a complex loss function within the SGD optimization process, replacing traditional amplitude-only loss functions.
• This complex loss function allows for a single total loss calculation, drastically cutting down computation time.
• Simultaneously optimized both amplitude and phase components of the object for more accurate complex amplitude distribution reconstruction.

Main Results:

• Achieved a substantial reduction in optimization time for multi-depth hologram generation compared to existing SGD methods.
• Enabled the accurate reconstruction of complex amplitude distribution, leading to a better match with the defocus blur effect.
• Validated the proposed method's effectiveness through both numerical simulations and physical optical experiments.

Conclusions:

• The proposed complex loss function method offers a computationally efficient and effective solution for SGD-based hologram optimization.
• This advancement makes practical the generation of high-quality holographic displays for complex, multi-depth 3D objects.
• The method enhances holographic display fidelity by accurately reconstructing complex amplitude distributions.