[Paper Review] Compression of phase-only holograms with JPEG standard and deep learning

Shuming Jiao, Zhi Jin+4 more|arXiv (Cornell University)|Jun 11, 2018

Advanced Optical Imaging Technologies35 references3 citations

TL;DR

This paper proposes a hybrid compression framework combining JPEG standard and deep learning to efficiently compress computer-generated phase-only holograms. By using a deep convolutional neural network to restore artifacts introduced by JPEG compression, the method achieves high-quality holographic reconstructions while maintaining compatibility with widely deployed JPEG infrastructure, significantly improving visual fidelity over standard JPEG-only compression.

ABSTRACT

It is a critical issue to reduce the enormous amount of data in the processing, storage and transmission of a hologram in digital format. In photograph compression, the JPEG standard is commonly supported by almost every system and device. It will be favorable if JPEG standard is applicable to hologram compression, with advantages of universal compatibility. However, the reconstructed image from a JPEG compressed hologram suffers from severe quality degradation since some high frequency features in the hologram will be lost during the compression process. In this work, we employ a deep convolutional neural network to reduce the artifacts in a JPEG compressed hologram. Simulation and experimental results reveal that our proposed "JPEG + deep learning" hologram compression scheme can achieve satisfactory reconstruction results for a computer-generated phase-only hologram after compression.

Motivation & Objective

• To address the challenge of high data volume in digital holography, particularly for computer-generated phase-only holograms.
• To enable universal compatibility by adapting the widely used JPEG standard for hologram compression.
• To mitigate severe image quality degradation caused by lossy JPEG compression of holographic data.
• To develop a deep learning-based restoration method that enhances reconstructed holographic quality after JPEG compression.
• To demonstrate the feasibility and effectiveness of a 'JPEG + deep learning' pipeline for practical hologram compression.

Proposed method

• Applying the standard JPEG compression algorithm to phase-only holograms to reduce data size.
• Using a deep convolutional neural network (CNN) to learn and remove compression artifacts from JPEG-compressed holograms.
• Training the CNN on pairs of original and JPEG-compressed holograms to learn the mapping from compressed to high-quality reconstructed images.
• Employing a U-Net-like architecture for end-to-end feature learning and artifact suppression.
• Optimizing the network using a loss function that minimizes the difference between reconstructed and original holographic intensity patterns.
• Integrating the deep learning model as a post-processing step after JPEG compression to restore image quality.

Experimental results

Research questions

• RQ1 Can the JPEG standard be effectively applied to compress phase-only holograms despite their high-frequency content?
• RQ2 To what extent does JPEG compression degrade the quality of reconstructed holographic images?
• RQ3 Can a deep learning model effectively restore visual quality after JPEG compression of holograms?
• RQ4 How does the proposed 'JPEG + deep learning' pipeline compare to JPEG-only compression in terms of reconstruction fidelity?
• RQ5 Is the proposed method compatible with existing systems and infrastructure designed for JPEG?

Key findings

• The proposed 'JPEG + deep learning' method significantly reduces visual artifacts in JPEG-compressed phase-only holograms.
• Simulation and experimental results confirm that the deep learning model effectively restores high-frequency features lost during JPEG compression.
• The method achieves perceptually superior reconstruction quality compared to JPEG-only compression, with improved signal-to-noise ratio and structural similarity.
• The approach maintains compatibility with existing JPEG infrastructure, enabling broad deployment potential.
• The deep learning component successfully learns to reconstruct fine details and textures in the holographic intensity patterns.
• The framework demonstrates robust performance across various holographic scenes and compression ratios.