Iterative approach for high-quality binary intensity hologram generation in augmented reality applications Journal: Optics and Laser in Engineering
Anik Ghosh, Matteo Ziliani, Marco Astarita, Samuele Trezzi, Alessandro Cerioni, Anna Cesaratto, Mara Galli, Gianluca Valentini, Andrea Bassi, Giulio Cerullo, Paolo Pozzi* · 2026
Read Iterative approach for high-quality binary intensity hologram generation in augmented reality applications Journal: Optics and Laser in Engineering by Anik Ghosh, Matteo Ziliani, Marco Astarita, Samuele Trezzi, Alessandro Cerioni, Anna Cesaratto, Mara Galli, Gianluca Valentini, Andrea Bassi, Giulio Cerullo, Paolo Pozzi* (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-11181.txt.
Introduction
Ghosh and ten colleagues tackle holograms made with binary light modulators, whose on-off pixels normally hurt image quality. Instead of binarizing a finished hologram, their step-by-step method enforces the binary rule at each round. Simulations and a mirror-device test show sharper contrast, better light use, and truer reconstructions than random superposition and Gerchberg-Saxton, at similar cost.
Catalog record
| CREATOR | Anik Ghosh, Matteo Ziliani, Marco Astarita, Samuele Trezzi, Alessandro Cerioni, Anna Cesaratto, Mara Galli, Gianluca Valentini, Andrea Bassi, Giulio Cerullo, Paolo Pozzi* |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34326 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.11181
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Iterative approach for high-quality binary intensity hologram generation in augmented reality applications Journal: Optics and Laser in Engineering Anik Ghosh, Matteo Ziliani, Marco Astarita, Samuele Trezzi, Alessandro Cerioni, Anna Cesaratto, Mara Galli, Gianluca Valentini, Andrea Bassi, Giulio Cerullo, Paolo Pozzi* Abstract Binary amplitude spatial light modulators, such as digital micromirror devices (DMDs), are increasingly relevant for computer-generated holography due to their high refresh rates, low cost, and due to the emergence of sub-wavelength pixel architectures. However, the binary constraint limits the reconstruction quality, as conventional approaches rely on a binarization applied as a final step after hologram computation which leads to reduced efficiency and contrast. We introduce an iterative estimation approach for the generation of off-axis binary amplitude holograms, in which the binarization constraint is applied at each iteration. We validate the approach through numerical simulations and experimental reconstruction using a DMD-based optical setup. Quantitative and qualitative comparisons with random superposition and Gerchberg–Saxton methods demonstrate significant improvements in image contrast, light efficiency, and reconstruction fidelity, with comparable computational cost. The proposed method provides a practical route toward high-quality CGH using binary modulators and supports emerging applications requiring high-speed and high-resolution holographic projection. 1 Introduction Computer-generated holography (CGH), first introduced by Brown and Lohmann in 1966 [ 8 ] , has since evolved into a robust method for precise optical wavefront control and high-fidelity image projection [ 35 , 6 ] . With the advancement of computational methods and spatial light modulators (SLMs), CGH is now widely adopted in various domains, including optical trapping [ 1 ] and optogenetics [ 9 , 15 , 22 ] . Moreover, it is considered as a promising new technology for the future of augmented reality (AR) systems [ 44 , 20 ] . Holography is a 2 step process, recording and reconstruction. In CGH, the first step is purely computational, while the second step is optical. However, computational demand increases with the complexity of the scene. Although holograms are inherently complex-valued, physical devices are not capable of modulating both amplitude and phase simultaneously. Instead, existing light modulators, such as Liquid Crystal on Silicon (LCOS) SLMs or digital micromirror devices (DMDs) are constrained to either phase-only or amplitude-only modulation, which inevitably leads to reduced image fidelity in reconstructions. In recent years, substantial progress has been made in CGH techniques, particularly those focusing on phase-only modulation, because of their superior diffraction efficiency. The Gerchberg–Saxton (GS) algorithm [ 16 ] , a widely used iterative approach, has been adapted to handle 2D [ 14 ] , multiplane scenes [ 41 , 28 , 34 ] and 3D point clouds [ 36 ] . An improved phase-CGH method was developed based on stochastic gradient descent (SGD) optimization [ 10 ] , although these methods are computationally intensive due to the involvement of complex gradient evaluations. Recently, a wireframe holography method for fast CGH generation was reported [ 3 ] . In addition to iterative methods, non-iterative approaches have been explored for phase-only hologram generation. Double-phase encoding, for instance, translates complex-valued holograms into two interleaved phase-only maps, preserving quality at the cost of halved spatial resolution [ 2 ] . Other techniques involve optimizing carrier waves, such as using random or quadratic phase patterns, to tailor the illumination for better phase-only reconstruction, although their effectiveness depends on the scene content [ 42 , 11 ] . Recently, deep learning has emerged as a fast and accurate alternative [ 43 ] , requiring large datasets for training but enabling high-quality hologram generation with minimal computation time. Incorporating experimental setup characteristics through camera-in-the-loop training [ 33 ] , deep learning has led to some of the best results in practical holography. High-performance phase-only SLMs are typically LCOS devices. However, LCOS-based devices have several drawbacks, preventing the widespread adoption of CGH in industrial and consumer application. Namely, while the high cost of the devices could be reduced in the future exploiting economies of scale, the large pixel sizes and low pixel refresh rate limit their applicability in fields such as AR displays, where sub-wavelength pixel sizes are required for large field of view projection, and kHz scale refresh rates are required for the suppression of speckle and speckle-like artifacts [ 38 ] . The refresh rate limitations of LCOS can already be inexpensively overcome through the use of DMDs [ 40 ] , microelectromechanic systems (MEMS) which perform binary amplitude modulation of the incoming light by switching micro-mirrors between two fixed positions. DMDs are capable of operating at frame rates up to tens of kilohertz and are free from pixel crosstalk. Additionally, they are significantly more cost-effective than LCOS-based SLMs. As a result, DMDs are now employed in a variety of fields including video projection [ 32 ] , fringe pattern generation [ 24 , 29 ] , beam shaping [ 4 ] , and holography [ 18 , 21 , 12 , 40 ] . Moreover, the need for nanoscale-size pixels in light modulators has led to the development of extremely high resolution modulators [ 23 ] , which are however currently only capable of binary amplitude modulation. This makes computation of binary amplitude hologram (BAH) critical for future implementations of CGH. However, generating BAH presents notable challenges. The use of only two amplitude levels inherently introduces significant quantization errors, leading to lower reconstruction quality. Early approaches [ 27 ] involved direct binarization of phase holograms, which caused substantial image degradation due to quantization artifacts. To alleviate this, researchers have pursued two main strategies: (i) refining binarization techniques and [ 19 ] (ii) optimizing the original phase hologram to partially retain quality after binarization [ 31 , 19 ] . These methods, while computationally fast, often compromise on image quality. An early solution to improve BAH quality involved binary search-based iterative algorithms [ 37 ] . However, these methods are computationally expensive and unsuitable for real-time use. Error diffusion techniques have also been applied [ 39 , 5 ] , offering reduced processing time but introducing high-frequency noise that restricts the resolution and scene complexity. A more recent advancement is the binarized stochastic gradient descent (B-SGD) method proposed by Lee et al. [ 26 ] , which minimizes reconstruction loss by optimization before binarization, achieving a better balance between quality and speed. In this work, we propose a novel method for the generation of high-quality BAHs using an iterative estimation framework integrated with intensity modulation and pre-calibrated system aberrations [ 17 , 30 , 6 ] . Iterative estimation is able to provide high-quality, low noise intensity holograms. Our approach embeds the binarization process within the iterative loop, resulting in an optimized binary amplitude output. To validate the effectiveness of the proposed technique, we implemented an off-axis holographic setup using a DMD and performed experimental reconstructions, confirming both its feasibility and high reconstruction quality, with an enhancement in image contrast of more than a factor 2 when compared with regular GS optimization. 2 Methodology 2.1 Binary amplitude hologram generation using iterative estimation method Figure 1: Schematic of the experimental setup for the reconstruction of binary holograms using a Digital Micromirror Device (DMD). A laser light source (LS) is used for illumination. The beam diameter is controlled and intensity-adjusted using an iris