With the continuous evolution of mobile device technologies, the integration of under‐display cameras has emerged as a groundbreaking innovation in the pursuit of bezel‐less and immersive visual experiences. As this technology becomes increasingly prevalent, it introduces a unique set of challenges, particularly in the realm of photography. In particular, the flare cannot be restored by conventional image processing due to saturated pixels. Therefore, a deep learning network capable of restoring such deterioration is proposed. The network is trained by a synthetic datasets generated by accurate optical simulation. However, the distortion of the camera lens distorts the shape of flare differently depending on its position in the photo, and it degrades restore quality. In this paper, we propose space‐variant CNN for solving distortion problem.

Under‐display camera (UDC) is a new form factor that implements a full‐screen display by locating the camera under the display panel. UDC suffer from image degradation and flare due to the panel in front of the camera. In particular, flare cannot be restored by conventional image processing due to the saturated pixels. Therefore, deep learning network that can restore these degradations is proposed. The network is trained by synthetic dataset generated by accurate optical simulation. However, the deep learning networks might not be suitable for video calls that require real‐time image processing due to their large amount of computation. The guided filter methodology is adapted to the network to reduce the amount of computation. Total amount of calculation is reduce by 90% and achieved 30 FPS at the FHD (1920×1080×3) resolution. In this paper, the method to generate realistic synthetic dataset that can solve the problems occurred during video restoration.

Under‐display camera suffers from severe image blur and glare due to the structure of the display panel. Finite dynamic range of camera also causes irreversible damage on images. In this paper, we show that the blurry with glare can be rapidly restored by a neural network with guided filters.

A commercial head-mounted display (HMD) for virtual reality (VR) presents three-dimensional imagery with a fixed focal distance. The VR HMD with a fixed focus can cause visual discomfort to an observer. In this article, we propose a novel design of a compact VR HMD supporting near-correct focus cues over a wide depth of field (from 18 cm to optical infinity). The proposed HMD consists of a low-resolution binary backlight, a liquid crystal display panel, and focus-tunable lenses. In the proposed system, the backlight locally illuminates the display panel that is floated by the focus-tunable lens at a specific distance. The illumination moment and the focus-tunable lens' focal power are synchronized to generate focal blocks at the desired distances. The distance of each focal block is determined by depth information of three-dimensional imagery to provide near-correct focus cues. We evaluate the focus cue fidelity of the proposed system considering the fill factor and resolution of the backlight. Finally, we verify the display performance with experimental results.

Key of development for OLED devices is compatibility between organic materials. Therefore, the construction of the materials’ properties DB should be followed in order to advance characterization and prediction of the OLED device. In particular, it is necessary to build a number of organic materials DB unlike the conventional Si‐based homo‐junction and alloy changing only heterojunction. So, it should be built based on the organic material parameters, SIMDB (Simulation DB), it is possible to predict and optimize driving voltage and efficiency characteristics for OLED devices. However, the skilled Engineer gets to decide on the simple repeat operation and time consumption for calibrating the material parameter/structure (4~5 layer device based on 20,000 times or more). In order to solve such problems, this paper presents an automatic optimization for electro‐optical characteristics of OLED device. The proposed method and tool are based on a unique heuristic algorithm, which is obtained from accumulated manual calibration experience. In other words, the convergence know‐how is applied random search algorithm and heuristic algorithm were constructed in‐house optimizer enables automated extraction of OLED properties. Finally, by utilizing the built SIMDB is expected to various factors affecting the OLED elements is optimized to be analyzed and prior prediction.

Recent developments in ultra-high-definition (UHD) displays have a major impact on holographic displays as well as conventional display systems. A spatial light modulator (SLM) can reproduce hologram data through wavefront modulation of the incident light, and a holographic printer can record a high-quality hologram by sequentially recording a number of holograms reproduced by the SLM on a holographic material. As UHD (4K) resolution SLMs have been popularized, higher quality holograms can be reproduced compared to previous 2K resolution SLMs. When applied to the holographic printer, it is possible to manufacture holographic optical elements (HOEs) having a multifunctioning property and a wide field of view. In this paper, we introduce the holographic printer system using an amplitude UHD SLM and its applications. The holographic printer consists of optical systems to generate high-quality hologram from UHD SLM and mechanical systems to record the hologram on holographic material. The details of the total system are introduced. Furthermore, we introduce a holographic near-eye display system using UHD SLM and an image combiner HOE which is manufactured by the holographic printer.

We present a deep neural network for generating a multi-depth hologram and its training strategy. The proposed network takes multiple images of different depths as inputs and calculates the complex hologram as an output, which reconstructs each input image at the corresponding depth. We design a structure of the proposed network and develop the dataset compositing method to train the network effectively. The dataset consists of multiple input intensity profiles and their propagated holograms. Rather than simply training random speckle images and their propagated holograms, we generate the training dataset by adjusting the density of the random dots or combining basic shapes to the dataset such as a circle. The proposed dataset composition method improves the quality of reconstructed images by the holograms generated by the network, called deep learning holograms (DLHs). To verify the proposed method, we numerically and optically reconstruct the DLHs. The results confirmed that the DLHs can reconstruct clear images at multiple depths similar to conventional multi-depth computer-generated holograms. To evaluate the performance of the DLH quantitatively, we compute the peak signal-to-noise ratio of the reconstructed images and analyze the reconstructed intensity patterns with various methods.

We propose a method to generate holograms using a deep neural network. The proposed network can generate complex holograms from slice images of different depths. It is verified that the images reconstructed from the hologram can be formed at two depths.

Intermediate pupil mask is adopted for synthesis of computer-generated holograms via Fourier ptychographic method. The problem of the conventional Fourier ptychographic method is that the exact target images cannot be created, when it utilized to synthesize computer-generated holograms. In the previous study, the diffraction effect on target images is ignored, and the target images are generated by light-field scheme. However, the mismatched target images degrade the quality of holographic images. The degradation can be mitigated by choosing the optimal size of the sub-regions. The optimal size of the sub-regions can be derived by considering the all-in-focus condition. It is also shown that the computation load can be reduced by setting the pupil mask to Gaussian function. The proposed method has an advantage in computation load for their image quality compared with other hologram synthesis algorithms. The proposed method is verified by numerical simulation.

In this paper, we propose computer-generated hologram calculation method for pentile display panel. In order to compensate non-orthogonal pixel struchrre, the CGHs for the even and odd rows of pixels are calculated separately and combined again with proper compensation term. The proposed method is verified by simulation and experimental results.

Speckle reduction is an important topic in holographic displays as speckles not only reduce signal-to-noise ratio but also possess an eye-safety issue. Despite thorough exploration of speckle reduction methods using partially coherent light sources, the trade-off involved by the partial coherence has not been thoroughly discussed. Here, we introduce theoretical models that quantify the effects of partial coherence on the resolution and the speckle contrast. The theoretical models allow us to find an optimal light source that maximizes the speckle reduction while minimizing the decline of the other terms. We implement benchtop prototypes of partially coherent holographic displays using the optimal light source, and verify the theoretical models via simulation and experiment. We also present a criterion to evaluate the depth of field in partially coherent holographic displays. We conclude with a discussion about approximations and limitations inherent in the theoretical models.

Multifocal displays which physically float multiple focal planes are one of the promising solutions to provide focus cues for near-eye displays. With respect to the focal plane arrangements, several works proposed to dynamically change focal plane configuration as it could efficiently cover wide depth range with a few focal planes to reduce focusing error. Although they optimize the plane locations based on the resultant retinal images for target volumetric scenes, computational loads to synthesize retinal images become significantly larger as the resolution of target contents increases. Here, we propose to exploit the deep neural network to figure out the optimized focal plane configuration without directly depicting resultant retinal images. We demonstrate that designed network computes the focal plane positions to achieve optimal retinal images through numerical simulations.

Latent fingerprints found at a crime scene are an important evidence in a criminal investigation. However, latent fingerprints are often damaged by the surface of an object or obscured by other latent fingerprints. Herein, we propose an end-to-end overlapped fingerprint separation method using a deep learning algorithm. The neural network, FinSNet, takes an overlapped fingerprint image as an input and is trained to restore the component fingerprint from the input directly. Our network can remove redundant fingerprints and background images with only a single inference, resulting in improved computational efficiency compared to conventional methods that require background removal to be preceded separately. First, we introduce strategies to design the network and build a training dataset. Next, we demonstrate results of separating real-world fingerprints, using Tsinghua overlapped fingerprint datasets. We then evaluate our proposed method using commercial fingerprint identification software to perform fingerprint matching. Finally, we thoroughly analyze our method in terms of the training dataset, background removal, and angular deviation of overlapped fingerprints.

Computational accommodation-invariant (AI) display attempts to mitigate vergence-accommodation conflict (VAC) by showing a constant imagery no matter where the observer focuses on. However, due to the usage of an electrically focus-tunable lens, the contrast of imagery is degraded as point-spread functions of multiple foci are integrated. In this paper, we introduce the content-adaptive approach to improve the contrast at the depth of highly salient region in the image. The position of focal plane is dynamically determined considering the zone of comfort and the mean focal distance of salient region. The contrast enhancement compared to conventional accommodation-invariant display is shown through simulation results using USAF resolution target image. We demonstrate our proof-of-concept prototype and its optical feasibility is verified with experimental results.

The ultimate 3D displays should provide both psychological and physiological cues for depth recognition. However, it has been challenging to satisfy the essential features without making sacrifices in the resolution, frame rate, and eye box. Here, we present a tomographic near-eye display that supports a wide depth of field, quasi-continuous accommodation, omnidirectional motion parallax, preserved resolution, full frame, and moderate field of view within a sufficient eye box. The tomographic display consists of focus-tunable optics, a display panel, and a fast spatially adjustable backlight. The synchronization of the focus-tunable optics and the backlight enables the display panel to express the depth information. We implement a benchtop prototype near-eye display, which is the most promising application of tomographic displays. We conclude with a detailed analysis and thorough discussion of the display's optimal volumetric reconstruction. of tomographic displays.

Light emitting diode (LED) has been a prominent component for illumination source of holographic displays. It can replace laser in that it alleviates speckle noise in reconstructed hologram and is regarded as a safer source. However, LED-based holographic displays can reconstruct an image in a limited depth range due to partially coherent characteristic of LED itself. In this paper, we propose a prototype of LED-based holographic near-eye display with focus tunable lens to expand the depth range. The feasibility of the system is supported by several experimental results.

Currently, commercial head-mounted displays suffer from limited accommodative states, which lead to vergenceaccommodation conflict. In this work, we newly design the architecture of head-mounted display supporting 15 focal planes over wide depth of field (20cm-optical infinity) in real time to alleviate vergence-accommodation conflict. Our system employs a low-resolution vertical scanning backlight, a display panel (e.g. liquid crystal panel), and focus-tunable lens. We demonstrate the compact prototype and verify its performance through experimental results.

Given the development of nano/microscale patterning techniques, efforts are being made to use them for fabricating metasurfaces. In particular, by using abrupt phase discontinuities, it is possible to generate holographic images from two-dimensional nanoscale-patterned metasurfaces. However, the fabrication of metasurface holograms is hindered by the high costs and long fabrication time involved, because the process requires expensive equipment such as that for electron-beam lithography. Therefore, it is difficult to realize metasurface holograms in a fast and repetitive manner. In this study, we propose a method for fabricating metasurface holograms based on the nanotransfer printing of the desired nanoscale patterns, which is assisted by Au nanoclusters, while controlling the bonding energy based on the shape of the deposited Au layer. Robust covalent bonds are formed between the Si of the adhesive used and the O of the SiO2 layer in order to transfer the deposited Au onto the transparent substrate quickly. It was found that the fabricated metasurface hologram coincides with the one designed by computer-generated holography. The proposed method should lead to a significant breakthrough in the fabrication of holograms based on different types of metasurfaces at a low cost in a fast, repetitive manner with various metals.

In this paper, an end-to-end optimization of optics and image processing which consider angle of incidences is proposed. By considering the various angle of incidences to the optics, the optimized system can capture and reconstruct a real image even for non-paraxial input light. The optimization pipeline includes diffractive wave simulation, effects from wavelength differences, and image processing. Several points spread functions are used to simulate captured images for tilted input light. Captured images are reconstructed by different deconvolution kernels according to the sub-section of the images. To apply the system to real-world experiments, we consider the limitation of diffraction angle for given memory constraints and differences between manufacturing and sensor resolution by using Fourier optics. We demonstrate the simulation results of the proposed approach by applying it to various angle of incidences.

In this paper, we propose a method to design and manufacture a holographic lens that can extend the eye-box in a near eye display. To verify the possibility, we calculated the phase profile of a holographic lens that produces multiple focal points with 5 mm intervals on the focal plane, and recorded it using a holographic wavefront printer. It was confirmed that the identical hologram images are displayed by spatial light modulator with 5 mm intervals on the focal plane of the manufactured holographic lens.