JSE0/VTM-FCH

---
license: other
license_name: vtm-fch-research-license-v1.0
license_link: LICENSE
size_categories:
- 100B<n<1T
---

# VTM-FCH (VTM-compressed full-complex hologram)

This dataset consists of full-complex holograms reconstructed after VTM compression and corresponds to the dataset used for training and testing in [our paper](). (To be updated)

## Introduction

Digital holography enables accurate representation of three-dimensional scenes by encoding both the amplitude and phase information of light waves. However, the extremely high spatial resolution of computer-generated holograms (CGHs) results in massive data volumes, making compression an essential step for storage and transmission. Conventional compression methods often introduce artifacts that significantly degrade the quality of numerical reconstruction (NR), particularly in perceptually important object regions.

The **VTM-FCH** dataset is designed to support research on compression, restoration, and quality assessment of full-complex holograms. This dataset contains holograms that have been compressed and reconstructed using the Versatile Video Coding Test Model (VTM) under standardized coding configurations.

A key characteristic of this dataset is that it reflects realistic compression degradations encountered in holographic data pipelines. The holograms originate from high-resolution CGH sources and are processed into non-overlapping patches suitable for deep learning–based training and evaluation.

By providing standardized VTM-compressed full-complex holograms, **VTM-FCH** serves as a benchmark dataset for advancing research in holographic signal processing and compression-aware hologram enhancement.

## Dataset Details and Structure

The **VTM-FCH** dataset includes 10 original holographic images. A summary of the **VTM-FCH** dataset is provided in the table below.

| Set | Type | CTC | Hologram | Original Resolution | Pixel Pitch | Wavelength |
|----------|------|-----|-----------------|---------------------|---------------------|---------------------|
| Training | CGH | ✓ | Dices16k | 16384×16384 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Training | CGH | ✓ | Ring16k | 16384×16384 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Training | CGH | ✓ | DeepDices16k | 16384×16384 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Training | CGH | ✓ | Biplane16k | 16384×16384 | 1e-6 | 640e-9, 532e-9, 473e-9 |
| Test | CGH | ✓ | Piano16k | 16384×16384 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Test | CGH | ✓ | SpecularCar16k | 16384×16384 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Test | OCH | ✓ | Lowiczanka Doll | 59394×2016 | 3.45e-6 | 637e-9, 532e-9, 457e-9 |
| Test | CGH | ✗ | Piano8k | 8192×8192 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Test | CGH | ✗ | Ring4k | 4096×4096 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |
| Test | CGH | ✗ | DiffuseCar4k | 4096×4096 | 0.4e-6 | 640e-9, 532e-9, 473e-9 |

All holograms are encoded using VTM-11.2 under the All-Intra configuration with YUV444 format.
Quantization parameters (QPs) are set to 22, 27, 32, 35, 38, and 41.
The holograms are partitioned into non-overlapping patches.
The dataset structure is as follows:

- **Training (66.1 GB)**
- Consists of patches with a resolution of 1024×1024, totaling 12,292 images.
- Includes *Dices16k*, *Ring16k*, *DeepDices16k*, and *Biplane16k*.
- **Test (46.2 GB)**
- Consists of patches with resolutions of 1024×1024 or 1024×1008, totaling 8,716 images.
- Includes *Piano16k*, *SpecularCar16k*, *Lowicznka Doll*, *Piano8k*, *Ring4k*, and *DiffuseCar4k*.

### Dataset Sources

We provide processed versions of the holographic images supplied by **B<>com**. The specifications of the original holograms are summarized below.

- **Repository:** [B<>com Hologram Repository](https://hologram-repository.labs.b-com.com/#/holographic-images)
- **Reference Papers:**
- A. Gilles, P. Gioia, R. Cozot, and L. Morin, "Hybrid approach for fast occlusion processing in computer-generated hologram calculation," Appl. Opt., AO, vol. 55, no. 20, pp. 5459-5470, Jul. 2016.
- A. Gilles, P. Gioia, R. Cozot, and L. Morin, "Computer generated hologram from Multiview-plus-Depth data considering specular reflections," in 2016 IEEE International Conference on Multimedia Expo Workshops (ICMEW), 2016, pp. 1-6.

### Copyright Notice

This dataset and resources are provided by the **Intelligent Visual Media Laboratory (IVML)** and are intended for **research and academic use only**.
By using this dataset, you agree to cite our paper "**ROI-guided Dual-domain Network for Enhanced Numerical Reconstruction of Compressed Holograms**" in any publications or presentations that utilize this dataset.

- **Copyright:** **Copyright (c) 2026 Intelligent Visual Media Laboratory (IVML)** The VTM-FCH dataset is copyrighted by the Intelligent Visual Media Laboratory (IVML). All rights are reserved. Unauthorized use, reproduction, or distribution of this dataset, or any portion thereof, is strictly prohibited without prior written permission from IVML. For any inquiries regarding the dataset, please contact us at [e-mail]. Permission is not granted, without written agreement and without license or royalty fees, to use, copy, modify, and distribute this dataset and its documentation for any purpose.
- **Acknowledgment:** This work was supported by the National Research Foundation of Korea (NRF) funded by Korean Government [Ministry of Science and ICT (MSIT)] under Grant RS-2023-00250751.

## Contact

For questions or requests regarding the dataset, please contact:

- Juyeon Seo, [e-mail]
- Hyeji Jang, [e-mail]