Phase-only hologram video compression using a deep neural network for up-scaling and restoration

Area of Science:

• Holography
• Digital signal processing
• Computer vision

Background:

• Phase-only hologram video (PoHV) compression is crucial for realistic 3D holographic displays.
• Existing video coding methods struggle with the low temporal correlation inherent in PoHV.
• Leveraging standard video codecs as anchors offers compatibility and accelerated development for new 3D data compression techniques.

Purpose of the Study:

• To propose an effective coding method for phase-only hologram video (PoHV) compression.
• To enhance the compression performance and visual quality of PoHV displayed on spatial light modulators.
• To overcome the challenge of low temporal correlation in PoHV using an anchor codec approach.

Main Methods:

• A novel coding method was developed using a standard video codec as an anchor.
• Progressive scaling and a deep neural network were integrated into the encoding and decoding processes.
• The method addresses low temporal correlation without increasing the anchor codec's complexity.

Main Results:

• The proposed coding method achieved an average coding gain of 22% compared to the anchor codec across all conditions.
• Numerical and optical reconstructions demonstrated clearer objects and reduced juddering in the results.
• The integration of scaling and neural networks effectively improved PoHV compression.

Conclusions:

• The developed coding method significantly enhances PoHV compression efficiency and visual fidelity.
• This anchor-based approach offers a viable solution for compressing 3D holographic data.
• The findings pave the way for improved real-time holographic display technologies.