角江 崇

We developed a HORN-8 system that generates computer-generated holograms at a high speed. The cluster system employed eight HORN-8 boards and achieved a level of performance that was 1,000 times faster than that of a PC. From a point-cloud model comprising 65,536 (216) points, the proposed cluster system can update a 2-million-pixel (1,920 × 1,080) hologram at 60 frames per second. 65,536 (216) is the internal memory size of the HORN-8 hardware. However, the HORN-8 system can calculate a hologram at a high speed even if the number of point-cloud sources exceeds 65,536 (216). Herein, we spatiotemporally divided a point-cloud model comprising ~400,000 points and succeeded in reproducing the video-holography. We demonstrated the performance of the special-purpose computer for electroholography using HORN-8 hardware that does not require a large internal memory when the calculation speed is high.

Electro-holography is a promising display technology that can reconstruct a photorealistic three-dimensional (3D) movie; however, it is yet to be realized practically owing to the need for enormous calculation power. A special-purpose computer for electro-holography, namely HORN, has been studied for over 20 years as a means to solve this problem. The latest version of HORN, HORN-8, was developed using field programmable gate array (FPGA) technology. Initially, a circuit for amplitude-type electro-holography was implemented in HORN-8; however, implementation of phase-type electro-holography has remained an issue. In this paper, the development of new version of HORN-8 and its cluster system, which achieved a real-time reconstruction of a 3D movie with point clouds comprised of 32,000 points for phase-type electro-holography, was reported.

© 2018 Chinese Optics Letters. We propose a method for color electroholography using a simple red–green–blue (RGB) gradation representation method without controlling the respective brightness of the reference RGB-colored lights. The proposed method uses RGB multiple bit planes comprising RGB binary-weighted computer-generated holograms with various light transmittances. The object points of a given three-dimensional (3D) object are assigned to RGB multiple bit planes according to their RGB gradation levels. The RGB multiple bit planes are sequentially displayed in a time-division-multiplexed manner. Consequently, the proposed method yields a color gradation representation of a reconstructed 3D object.

We propose a high-speed image-reconstruction algorithm for a spatially multiplexed image that is obtained by spatial frequency-division multiplexing. The algorithm utilizes smoothing and does not require any Fourier transform (FT) or iterative procedure to extract the desired information selectively from a single image. Numerical and experimental results show its validity and color holographic imaging ability. Calculation time of the proposed is less than a tenth of that of the FT method when using a central processing unit and an image sensor with four megapixels. Furthermore, throughput that is three times that of the FT method can be achieved in the case of using a graphics processing unit.

© 2018 Society of Photo-Optical Instrumentation Engineers (SPIE). Parallel phase-shifting digital holography is a powerful technique for recording motion picture of holograms with an image sensor and numerically reconstructing the motion picture of the high-quality three-dimensional (3-D) images of the object in a computer. This technique provides high-quality images of light intensity and phase of the object at a time instant at arbitrary depth position. Therefore, the technique achieves 3-D motion-picture imaging of a dynamic object. The technique provides not only large depth of field and high temporal resolution at the same time but also a motion picture of an invisible object. It is difficult for other 3-D imaging techniques to achieve these features of parallel phase-shifting digital holography. The authors review two experiments demonstrating the 3-D imaging by the technique and the features of the technique. One experiment demonstrated the motion-picture 3-D imaging of a minute crystal, sinking down in solution, by refocusing the amplitude images of the crystal. The result was obtained for the first time, to the best of our knowledge. The other demonstrated the motion-picture 3-D imaging of refractive indices of dynamic invisible gas flow by applying the Abel inversion to the phase images of the flow obtained by the technique.

© The Authors 2018. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Electroholography enables the projection of three-dimensional (3-D) images using a spatial-light modulator. The extreme computational complexity and load involved in generating a hologram make real-time production of holograms difficult. Many methods have been proposed to overcome this challenge and realize real-time reconstruction of 3-D motion pictures. We review two real-time reconstruction techniques for aerial-projection holographic displays. The first reduces the computational load required for a hologram by using an image-type computer-generated hologram (CGH) because an image-type CGH is generated from a 3-D object that is located on or close to the hologram plane. The other technique parallelizes CGH calculation via a graphics processing unit by exploiting the independence of each pixel in the holographic plane.

Holography is a method of recording and reproducing three-dimensional (3D) images, and the widespread availability of computers has encouraged the development of holographic 3D screens (electroholography). However, the technology has not yet been used in practical applications because a hologram requires an enormous volume of data and modern computing power is inadequate to process this volume of data in real time. Here, we show that a special-purpose holography computing board, which uses eight large-scale field-programmable gate arrays, can be used to generate 10(8)-pixel holograms that can be updated at a video frame rate. With our approach, we achieve a parallel operation of 4,480 hologram calculation circuits on a single board, and by clustering eight of these boards, we can increase the number of parallel calculations to 35,840. Using a 3D image composed of 7,877 points, we show that 108-pixel holograms can be updated at a video rate, thus allowing 3D movies to be projected. We also demonstrate that the system speed scales up in a linear manner as the number of parallel circuits is increased. The system operates at 0.25 GHz with an effective speed equivalent to 0.5 petaflops (10(15) floating-point operations per second), matching that of a high-performance computer.

In this paper, we demonstrate an interactive, finger-sensitive system which enables an observer to intuitively handle electro-holographic images in real time. In this system, a motion sensor detects finger gestures (swiping and pinching) and translates them into the rotation and enlargement/reduction of the holographic image, respectively. By parallelising the hologram calculation using a graphics processing unit, we realised the interactive handling of the holographic image in real time. In a demonstration of the system, we used a Leap Motion sensor and a phase modulation-type spatial light modulator with 1,920 × 1,080 pixels and a pixel pitch of 8.0 µm × 8.0 µm. The constructed interactive finger-sensitive system was able to rotate a holographic image composed of 4,096 point light sources using a swiping gesture and enlarge or reduce it using a pinching gesture in real time. The average calculation speed was 27.6 ms per hologram. Finally, we extended the constructed system to a full-colour reconstruction system that generates a more realistic three-dimensional image. The extended system successfully allowed the handling of a full-colour holographic image composed of 1,709 point light sources with a calculation speed of 22.6 ms per hologram.

Although electro-holography can reconstruct three-dimensional (3D) motion pictures, its computational cost is too heavy to allow for real-time reconstruction of 3D motion pictures. This study explores accelerating colour hologram generation using light-ray information on a ray-sampling (RS) plane with a graphics processing unit (GPU) to realise a real-time holographic display system. We refer to an image corresponding to light-ray information as an RS image. Colour holograms were generated from three RS images with resolutions of 2,048 × 2,048; 3,072 × 3,072 and 4,096 × 4,096 pixels. The computational results indicate that the generation of the colour holograms using multiple GPUs (NVIDIA Geforce GTX 1080) was approximately 300-500 times faster than those generated using a central processing unit. In addition, the results demonstrate that 3D motion pictures were successfully reconstructed from RS images of 3,072 × 3,072 pixels at approximately 15 frames per second using an electro-holographic reconstruction system in which colour holograms were generated from RS images in real time.

© 2018 International Display Workshops. All rights reserved. We report high-speed 3-D spatiotemporal division multiplexing holographic movie playback using the packed data of computer-generated hologram stored in high-performance solid state drive. Consequently, we succeeded to play high image quality 3-D movie of the 3-D object comprising about 900,000 points at 60 frames per second.

© 2018 International Display Workshops. All rights reserved. We demonstrated real-time electroholographic 3-D movie reconstruction using spatiotemporal division multiplexing technique on a multiple GPU cluster system including 13 GPUs connected through gigabit ethernet network. We succeeded to display reconstructed 3-D movie consisting of 477,511 object points at 36 fps. t.

© 2018 International Display Workshops. All rights reserved. We report real-time holographic projection to reconstruct 3D-gradation movie on cubic screen by using a multiple bit planes based on binary-weighted computer-generated holograms and a multiple-graphics-processing-unit cluster system. Consequently, we succeeded to reconstruct real-time 3D-gradation movie comprising 4840 points at 120 fps.

copyright © 2016 Society of Information Display. All rights reserved. We accelerated hologram generation based on raysampling plane by a graphics processing unit (GP U). The computational time by a central processing unit (CPU) was 56.02 seconds, while that by a GPU was 0.3764 seconds. We achieved to generate a 3072×3072-pixels hologram by the GPU approximately 150 times faster than the CPU.

copyright © 2016 Society of Information Display. All rights reserved. We propose high-speed computer-generated hologram reproduction using digital mirror device for high-definition spatiotemporal division multiplexing electro-holography. Finally, we succeeded to play high-definition 3-D movie of 3-D object comprised about 900,000 points at 60 fps when each frame was divided into twelve.

copyright © 2016 Society of Information Display. All rights reserved. We proposed a volumetric display composed of photochromic materials. It can be controlled in a non-contact from the outside. In this study, the photochromic material was mixed into transparent silicone to be solid. We made a volumetric display that can display different multiple pictures depending on the surface to observe. The gradation expression is realized by overlapping the colored voxels in a row to the observed directions.

© 2018 International Display Workshops. All rights reserved. By using the holographic optical element screen, an aerial-projection display of three-dimensional (3-D) images can be realized up close which the scale is free. However, the projected image is distorted when an object is placed far from the hologram plane. In this study, we corrected the distortion by shift point cloud.

© 2018 International Display Workshops. All rights reserved. In this study, we verify that the target observation area is shifted and expand viewing window by using off-axis reference light. We confirmed that the target observation area was shifted ±5 mm by changing the angle of off-axis reference from -1.5 degree to+1.5 degree.

© 2018 International Display Workshops. All rights reserved. For realizing electroholography, compact and highperformance computer is required. In this study, we implemented simple and fast cosine approximation method to create computer-generated hologram. As a result, we succeeded in reducing memory use rate 1/23 compared to using the look-up table of the cosine.

copyright © 2016 Society of Information Display. All rights reserved. The combination of the random phase-free method and Gerchberg-Saxton (GS) algorithm succeeded in improving the image quality of holograms. However, the GS algorithm takes a long computation time. In this research, we propose faster methods for the image quality improvement of the random phase-free hologram.

© 2018 International Display Workshops. All rights reserved. We reduced the computational load of hologram generation by extracting moving objects in real scene using the background subtraction to realize real-time electro-holographic display system. The hologram generation with the background subtraction was approximately 12 times faster than that without the background subtraction.

© 2018 by Society for Information Display. All rights reserved. We propose a calculation circuit for "Computational Ghost Imaging". Computational ghost imaging is a unique technique that allows single pixel imaging. However, this technique needs much calculation time. Therefore, we tried to shorten the calculation time by designing a circuit for computational ghost imaging and implementing it into an FPGA.

© 2018 IEEE. Digital holography enables us to reconstruct objects in three-dimensional space from holograms captured by an imaging device. For the reconstruction, we need to know the depth position of the recoded object in advance. In this study, we propose depth prediction using convolutional neural network (CNN)-based regression. In the previous researches, the depth of an object was estimated through reconstructed images at different depth positions from a hologram using a certain metric that indicates the most focused depth position; however, such a depth search is time-consuming. The CNN of the proposed method can directly predict the depth position with millimeter precision from holograms.

In this study, we describe the fabrication of a high-resolution directional volumetric display that can display multiple images in different directions. The display designs can be used to show animations using strings; however, improving the resolution of such displays is difficult. Previously, the arrangement of strings has only been determined experimentally, making fabrication of volumetric displays a challenge. The goal of the present study is to improve resolution using simulations and to determine the arrangement of strings under three constraints. This simplified the fabrication of a directional volumetric display with 345 strings, which can display two different 20×20  pixel images in two different directions. A large high-resolution directional volumetric display can be fabricated using the proposed method. The string-type display has high artistic potential and is expected to find applications in the amusement and entertainment fields.

© 2018 SPIE. Parallel phase-shifting digital holography is a technique capable of recording three-dimensional (3D) motion picture of dynamic object, quantitatively. This technique can record single hologram of an object with an image sensor having a phase-shift array device and reconstructs the instantaneous 3D image of the object with a computer. In this technique, a single hologram in which the multiple holograms required for phase-shifting digital holography are multiplexed by using space-division multiplexing technique pixel by pixel. We demonstrate 3D motion picture of dynamic and transparent gas flow recorded and reconstructed by the technique. A compressed air duster was used to generate the gas flow. A motion picture of the hologram of the gas flow was recorded at 180,000 frames/s by parallel phase-shifting digital holography. The phase motion picture of the gas flow was reconstructed from the motion picture of the hologram. The Abel inversion was applied to the phase motion picture and then the 3D motion picture of the gas flow was obtained.

We demonstrate fast time-division color electroholography using a multiple-graphics-processing-unit (GPU) cluster system with a spatial light modulator and a controller to switch the color of the reconstructing light. The controller comprises a universal serial bus module to drive the liquid crystal optical shutters. By using the controller, the computer-generated hologram (CGH) display node of the multiple-GPU cluster system synchronizes the display of the CGH with the color switching of the reconstructing light. Fast time-division color electroholography at 20 fps is realized for a three-dimensional object comprising 21,000 points per color when 13 GPUs are used in a multiple-GPU cluster system.

© 2017 The Institute of Electronics. This study involves proposing a high-speed computergenerated hologram playback by using a digital micromirror device for high-definition spatiotemporal division multiplexing electroholography. Consequently, the results indicated that the study successfully reconstructed a high-definition 3-D movie of 3-D objects that was comprised of approximately 900,000 points at 60 fps when each frame was divided into twelve parts.

We propose a novel algorithm that does not require any Fourier transform to extract multiple object waves in a single image recorded with spatial frequency-division multiplexing. Smoothing is utilized to extract the desired object-wave information from a spatially multiplexed image. Numerical and experimental results show its validity and applicability for image and Fresnel digital holography. Our investigations clarify the speeding up of both the object-wave extractions and multiple object-image reconstructions quantitatively. (C) 2017 Elsevier B.V. All rights reserved.

Computer-generated holograms (CGHs) are a key technology in electroholography systems; however, heavy calculations are required to calculate CGHs. We review fast calculation techniques for CGH calculation of a point-light-source-based model, which is a simple and general model of a three-dimensional object in an electroholography system. To reduce the calculation time, many methods that reduce the temporal and spatial redundancy of the CGH calculation have been developed (e.g., look-up table method, the wavefront recording plane method, and other approximation techniques). The implementation of such methods on parallel computers (e.g., graphic processing unit and field programmable gate arrays) has also been reported.

Digital holography allows production of high-speed three-dimensional images at rates over 100,000 frames per second; however, simultaneously obtaining suitable performance and levels of accuracy using digital holography is difficult. This problem prevents high-speed three-dimensional imaging from being used for vibrometry. In this paper, we propose and test a digital holography method that can produce vibration measurements. The method is based on single-shot phase-shifting interferometry. Herein, we imaged the surface of a loudspeaker diaphragm and measured its displacement due to the vibrations produced by a frequency sweep signal. We then analyzed the frequency of the experimental data and confirmed that the frequency spectra inferred from the reconstructed images agreed well with the spectra produced by the sound recorded by a microphone. This method can be used for measuring vibrations with three-dimensional imaging for loudspeakers, microelectromechanical systems, surface acoustic wave filters, and biological tissues and organs.

We propose a deep-learning- based classification of data pages used in holographic memory. We numerically investigated the classification performance of a conventional multilayer perceptron (MLP) and a deep neural network, under the condition that reconstructed page data are contaminated by some noise and are randomly laterally shifted. When data pages are randomly laterally shifted, the MLP was found to have a classification accuracy of 93.02%, whereas the deep neural network was able to classify data pages at an accuracy of 99.98%. The accuracy of the deep neural network is 2 orders of magnitude better than the MLP. (C) 2017 Optical Society of America

We designed and developed a control circuit for a three-dimensional (3-D) light-emitting diode (LED) array to be used in volumetric displays exhibiting full-color dynamic 3-D images. The circuit was implemented on a field-programmable gate array; therefore, pulse-width modulation, which requires high-speed processing, could be operated in real time. We experimentally evaluated the developed system by measuring the luminance of an LED with varying input and confirmed that the system works appropriately. In addition, we demonstrated that the volumetric display exhibits different full-color dynamic two-dimensional images in two orthogonal directions. Each of the exhibited images could be obtained only from the prescribed viewpoint. Such directional characteristics of the system are beneficial for applications, including digital signage, security systems, art, and amusement. (C) The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License.

Although digital holography is a powerful technique obtaining a phase image of a transparent object, the image reconstructed by the technique merely expresses phase distribution of the light wave after transmitting through the object. Phase variation of inside of the object is difficult to be obtained. Then, we applied Abel inversion method to the highspeed phase image of a dynamic transparent object assumed axially symmetric. The phase is accurately recorded by phase-shifting method. We experimentally recorded transparent dynamic gas flow, assumed axially symmetric along the direction in which gas flowed, at 3,000 frame/s and reconstructed motion picture of 3D distribution of the refractive index of the gas from the high-speed phase motion picture obtained by parallel phase-shifting digital holography. (C) 2017 Optical Society of America

Fast hologram calculation methods are critical in real-time holography applications such as three-dimensional (3D) displays. We recently proposed a wavelet transform-based hologram calculation called WASABI. Even though WASABI can decrease the calculation time of a hologram from a point cloud, it increases the calculation time with increasing propagation distance. We also proposed a wavefront recoding plane (WRP) method. This is a two-step fast hologram calculation in which the first step calculates the superposition of light waves emitted from a point cloud in a virtual plane, and the second step performs a diffraction calculation from the virtual plane to the hologram plane. A drawback of the WRP method is in the first step when the point cloud has a large number of object points and/or a long distribution in the depth direction. In this paper, we propose a method combining WASABI and the WRP method in which the drawbacks of each can be complementarily solved. Using a consumer CPU, the proposed method succeeded in performing a hologram calculation with 2048 x 2048 pixels from a 3D object with one million points in approximately 0.4 s.

We propose a simple gradation representation method using a binary-weighted computer-generated hologram (CGH) to be displayed on a high-speed spatial light modulator that can be controlled by the pulse-width modulation technique. The proposed method uses multiple bit planes comprising binary-weighted CGHs with various pulse widths. The object points of a three-dimensional (3D) object are assigned to multiple bit planes according to their gray levels. The bit planes are sequentially displayed in a time-division-multiplexed manner. Consequently, the proposed method realizes a gradation representation of a reconstructed 3D object.

We propose a holographic image restoration method using an autoencoder, which is an artificial neural network. Because holographic reconstructed images are often contaminated by direct light, conjugate light, and speckle noise, the discrimination of reconstructed images may be difficult. In this paper, we demonstrate the restoration of reconstructed images from holograms that record page data in holographic memory and quick response codes by using the proposed method. (C) 2017 Optical Society of America

A holographic projector utilizes holography techniques. However, there are several barriers to realizing holographic projections. One is deterioration of hologram image quality caused by speckle noise and ringing artifacts. The combination of the random phase-free method and the Gerchberg-Saxton (GS) algorithm has improved the image quality of holograms. However, the GS algorithm requires significant computation time. We propose faster methods for image quality improvement of random phase-free holograms using the characteristics of ringing artifacts. (C) 2017 Optical Society of America

Digital holography is expected to be useful in the analysis of moving three-dimensional (3D) image measurement. In this technique, a two-dimensional interference fringe recorded using a 3D image is captured with an image sensor, and the 3D image is reproduced on a computer. To obtain the reproduced 3D images with high spatial resolution, a high-performance image sensor is required, which increases the system cost. We propose an algorithm for super-resolution processing in digital holography that does not require a high-performance image sensor. The proposed algorithm wherein 3D images are considered as the aggregation of object points improves spatial resolution by performing a light-intensity search of the reproduced image and the object points.

In this study, a method to construct a full-colour volumetric display is presented using a commercially available inkjet printer. Photoreactive luminescence materials are minutely and automatically printed as the volume elements, and volumetric displays are constructed with high resolution using easy-to-fabricate means that exploit inkjet printing technologies. The results experimentally demonstrate the first prototype of an inkjet printing-based volumetric display composed of multiple layers of transparent films that yield a full-colour three-dimensional (3D) image. Moreover, we propose a design algorithm with 3D structures that provide multiple different 2D full-colour patterns when viewed from different directions and experimentally demonstrate prototypes. It is considered that these types of 3D volumetric structures and their fabrication methods based on widely deployed existing printing technologies can be utilised as novel information display devices and systems, including digital signage, media art, entertainment and security.

We propose a holographic microinformation hiding scheme in which the embedding information to be embedded is small and imperceptible to the human eyes. This scheme converts the embedding information into a complex amplitude via scaled diffraction. The complex amplitude of the reduced embedding information is added to the complex amplitude of the host image, followed by conversion to a hologram. The reduced embedded information is inconspicuous from the hologram during the reconstruction process; however, the reduction leads to the degradation of the embedded image quality. Therefore, to improve the quality of the embedded image quality, we employ iterative optimization and the time averaging effect of multiple holograms. (C) 2017 Optical Society of America

We propose a simple gradation representation method for a reconstructed three-dimensional (3-D) image without controlling the brightness of the reference light. In the proposed method, we use multiple bit planes comprised of binary-weighted computer-generated holograms (CGHs) with various light transmittances. Binary-weighted CGH is generated by changing the white in the conventional binary CGH to gray. The light transmittance of a binary-weighted CGH is less than that of a conventional binary CGH. The object points of a 3-D object are assigned to multiple bit planes according to the gray level of the object points. The multiple bit planes are displayed sequentially in a time-division multiplex manner. Consequently, the proposed method realizes a gradation representation of a reconstructed 3-D object. (C) 2017 Society of Photo-Optical Instrumentation Engineers (SPIE)

© 2017 Proceedings of the International Display Workshops. All rights reserved. We propose real-time electroholography to reconstruct 3D-gradation movie by using a multiple bit planes based on binary-weighted computer-generated holograms and a multiple-graphics-processing-unit cluster system. Consequently, we succeeded to reconstruct real-time 3D-gradation movie comprising 7850 points at 77 fps.

© 2017 Proceedings of the International Display Workshops. All rights reserved. We propose high-speed 3-D holographic movie playback including the reproduction of computergenerated holograms from the compression data stored in solid state drive. Consequently, we succeeded to playback high-definition movie of the 3-D object comprising about 900,000 points at 30 fps by using spatiotemporal division electroholography.

© 2017 Proceedings of the International Display Workshops. All rights reserved. For the practical use of real-time electroholography, we propose the multi-GPU cluster system with 13 GPUs (NVIDIA GeForce GTX TITAN X) and gigabit ether net network. Finally, we succeed to display reconstructed 3-D movie consisting of 95,949 object points at about 30 fps.

© 2017 Proceedings of the International Display Workshops. All rights reserved. We demonstrated real-time electro-holographic movie reconstruction using spatiotemporal division multiplexing technique on a GPU cluster including 13 GPUs connected through InfiniBand network. We succeeded to display reconstructed 3-D movie consisting of 477,511 object points at 31 fps.