Dynamic-range compression scheme for digital hologram using a deep neural network.

Area of Science:

• Digital holography
• Image processing
• Machine learning

Background:

• Digital holograms contain extensive data, posing storage and transmission challenges.
• Current compression methods often compromise the quality of reconstructed holographic images.
• Dynamic-range compression is crucial for efficient digital hologram handling.

Purpose of the Study:

• To propose a novel dynamic-range compression and decompression scheme for digital holograms.
• To leverage deep neural networks (DNNs) for high-fidelity holographic image reconstruction after compression.
• To enhance the efficiency of digital hologram storage and transmission.

Main Methods:

• A deep neural network (DNN) was employed for hologram decompression.
• A simple thresholding technique was used to compress 8-bit holograms to binary format.
• The error-diffusion algorithm was integrated into the binarization process to aid DNN training.
• The DNN was trained to predict original holograms from binary representations.

Main Results:

• The proposed scheme successfully compressed digital holograms to binary format, reducing data size by one-eighth.
• Despite initial quality degradation from binarization, the DNN effectively reconstructed high-quality images.
• The DNN-based scheme demonstrated superior performance compared to JPEG 2000 and high-efficiency video coding.

Conclusions:

• The proposed DNN-based dynamic-range compression scheme offers an effective solution for digital holograms.
• This approach significantly reduces data requirements while maintaining excellent reconstructed image quality.
• The method shows potential for practical applications in digital holographic data management.