This PDF file contains the front matter associated with SPIE Proceedings Volume 11353 including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

Computer-generated holography (CGH) has demonstrated significant promise as a display technique, with benefits including the ability to display images supporting true refocus at diffraction-limited resolution at the eye, and brightness levels meeting the most demanding requirements for outdoor augmented-reality applications. Whereas historically the computational load required for CGH has been a barrier to wider adoption, this has been demonstrated to be within reach of modern mobile devices for many real-world use cases. However, one area which remains a concern for CGH, as with other laser-based display technologies, is the reduction of noise arising from coherent illumination, loosely known as laser speckle. In this work we identify and model the modes in which Speckle affects the apparent visual quality of images generated using 2D and 3D holography and compare the effectiveness of known and theoretical optical and algorithmic techniques for speckle reduction. Reduction of speckle noise is of particular interest to CGH, but there is significant extant literature on speckle-reduction techniques from other disciplines, such as microscopy using coherent illumination, and laser metrology. These techniques are examined and their applicability to CGH considered. Approaches including a diffusing element are typically limited to 2D CGH, as they destroy the coherence required to support multiple depth planes. For techniques that in a broad sense act on the spatial/temporal coherence of the system to reduce speckle, such as use of a superluminescent light emitting diode (SLED), or a moving optical element, a trade-off between speckle reduction and reduced resolution must be found. In addition, we review the impact of algorithmic techniques used in CGH to reduce speckle-like noise during the generation of holograms, and compare the effectiveness of these techniques to hardware-based solutions.

Computer generated holography (CGH) is a computationally intensive task, which requires the design of efficient algorithms for calculating realistic holograms in real-time. Sparse CGH algorithms address this challenge by computing holographic signals in a transform domain where they are sparse, thereby requiring only to update a small number of coefficients and speeding up calculations. However, the required memory access patterns form a bottleneck, limiting speed gains. We propose a novel technique for computing phase-added stereograms by grouping Fourier coefficients into sub-stereograms, mapping to contiguous cells in 3D space. We present various implementations of this algorithm, reporting speedup factors ranging from 3 to 30 depending on the devices’ computing capability, without compromising on quality. This enables real-time CGH for driving holographic displays for more complex scenes.

Due to its ability to reproduce the correct focus cues, holography is considered as a promising display technology for Augmented Reality glasses. However, since it contains a large amount of data, the calculation of a hologram is a time-demanding process, resulting in prohibiting head-motion-to-photon latency. In this paper, we propose a real-time hologram calculation method based on two modules: an offline pre-computation module and an on-the-fly hologram synthesis module. In the offline calculation module, the omnidirectional light field scattered by each scene object is individually pre-computed and stored in a Look-Up Table (LUT). Then, in the hologram synthesis module, the light waves corresponding to the viewer’s position and orientation are extracted from the LUT in real-time to compute the hologram. Contrarily to previously proposed methods, our approach handles several independent scene objects and arbitrary user positions and orientations. Experimental results show that the proposed method is able to compute full-HD holograms at more than 256 frames per second, enabling its use in Augmented Reality applications.

Creating computer-generated holograms (CGHs) is computationally costly, and many high-speed calculation algorithms have been proposed to address this problem. Recently, a calculation method using sparse point spread functions (PSFs) in the short-time Fourier transform (STFT) domain has been proposed. Since the PSFs are sparse in the STFT domain, only a small fraction of STFT coefficients need to be calculated. Further, in order to obtain the STFT coefficient, fast Fourier transform is generally necessary, but this new method can obtain the STFT coefficient by analytical means. This means that the number of calculations required can be greatly reduced. This paper describes the implementation of an STFTbased CGH calculation algorithm on a field-programmable gate array. All operations in this algorithm were implemented using fixed point arithmetic. Since this algorithm includes a trigonometric function and an error function, we used look-up tables (LUTs) to reduce calculation costs. We have devised a dedicated circuit architecture that allows parallel operations. As a secondary effect, a central processing unit was able to generate holograms using the STFT-based CGH calculation algorithm with fixed point arithmetic and LUTs, faster than by using floating point arithmetic.

The possibility of employing the spatial degree of photons for communications is gaining interest in recent years due to its unbounded dimensionality. A natural basis to span the transverse profile of photons is comprised of Laguerre-Gaussian (LG) modes which are characterized with two topological numbers: l, the orbital index, describing the orbital-angular-momentum (OAM) in units of “h-bar” per photon in the beam and p, which is the radial index or radial quantum number. One of the main challenges for utilizing LG modes in communications is the ability to perform mode-sorting and demultiplexing of the incoming physical data-flow. Nowadays, there are two leading approaches to mode demultiplexing. The first approach uses intricate optical setups in which the l and p degrees of freedom are coupled to other degrees of freedom such as the angle of propagation or the polarization of the beam. Most of these methods address either the OAM or the radial index degrees of freedom. The second approach, which emerged recently, suggests using just a camera to detect the intensity of the incoming light beam and to utilize a deep neural network (DNN) to classify the beam. To date, demonstrated DNN-based demultiplexers addressed solely the OAM degree of light. We report on an experimental demonstration of state-of-the-art mode demultiplexing of Laguerre-Gaussian beams according to both their orbital angular momentum and radial topological numbers using a flow of two concatenated deep neural networks. The first network serves as a transfer function from experimentally-generated to ideal numerically-generated data, while using a unique "Histogram Weighted Loss" function that solves the problem of images with limited significant information. The second network acts as a spatial-modes classifier. Our method uses only the intensity profile of modes or their superposition with no need for any phase information.

Active contour-based methods are widely popular in the image segmentation field. Basically, they perform a semiautomatic region identification by partitioning the image content mainly into the foreground and background. Nevertheless, the accurate delimitation still remains as an important challenge which usually depends on how close the initial contour is placed to the object of interest (OI). Several applications of active contours require the user interaction to give prior information about the initial position as the first step, which drives the tool substantially dependent on a manual process. This paper describes how to overcome this limitation by including the expertise provided by the training stage of a Convolutional Neural Network (CNN). Despite CNN methods require a large dataset or data augmentation techniques to improve their results, the combined proposal accomplishes a presegmentation task with a reduced number of images to obtain the assumed locations for each OI. These results are used to initialize a multiphase active contour model that follows a level set scheme to lead a smoother multiregion segmentation with less effort. Experiments of this approach are included to compare classic techniques of contour initialization and show the benefits of our proposal.

Although the many advantages of Interventional Radiology not only being a minimally invasive surgery but also providing minimal risk of infection and better recuperation for the patient. However, this procedure can cause serious damage (cancer or burnt skin) to the patient and especially to the surgeons if they are exposed for long periods to the X-ray radiation. In the state of art, it has been found new remote catheter navigation system in which the equipment uses magnetic fields for controlling and moving the catheter from an external cabin. Additionally, large equipment needs to be installed in the operating room. In order to limit the doses of X-rays without installing large equipment in the operating room, the aim is to decrease the images coming from X rays imagers using sensors that can be integrated into the catheter (like Fibber Bragg Gratings Sensors inside the catheter or MEMS sensors) to reconstruct an image without the need of continuous imaging. In order to do that, accurate and reliable information on the position of the catheter is required to correct the drift of the catheter's sensors. This position can therefore be obtained by image processing on X-ray images (noisy with artefacts). Previous work done by the Medic@ team has shown that conventional image processing approaches are generally too slow or not precise enough. The use of a U-Net convolutional neural network is then a possible solution for detecting the entire catheter (body and head) and obtaining the coordinates of its end. In this article, we will explain and show our first results using the U-net architecture for detecting the tip and the body of the catheter and a kalman filter used for data fusion to evaluate its efficiency to reducing the quantity of images needed in a curvilinear vessel, using generated data.

Objects without prominent textures pose challenges for an automatic 3D model reconstruction and feature point matching. Such objects are common in many industrial applications such as metal defect detection, archeological applications of photogrammetry, and 3D object reconstruction from infrared imagery. Most of the common feature point descriptors fail to match local patches in featureless regions of an object. Various kind of textures requires different feature descriptors for high-quality image matching. Hence, automatic low-textured 3D object reconstruction using Structure from Motion (SfM) methods is challenging. Nevertheless, such reconstruction is possible with the aid of a human operator. Deep learning-based descriptors have outperformed most of common feature point descriptors recently. This paper is focused on the development of a new conditional generative adversarial auto-encoder (GANcoder) based on the deep learning. We use a coder-decoder architecture with four convolutional and four deconvolutional layers as a staring point for our research. Our main contribution is a generative adversarial framework GANcoder for training the auto-encoder on the textureless data. Traditional training approaches using an L1 norm tend to converge to the mean image on the low-textured images. In contrast, we use an adversarial discriminator to provided an additional loss function that is focused on distinguishing real images from the training dataset from the auto-encoder reconstruction. We collected a large GANPatches dataset of feature points from nearly textureless objects to train and evaluate our model and baselines. The dataset includes 16k pairs of image patches. We performed qualitative evaluation of our GANcoder and baselines for two tasks. Firstly, we compare the matching score of the our GANcoder and baselines. Secondly, we evaluate the accuracy of 3D reconstruction of low-textured objects using an SfM pipeline with stereo-matching provided by our GANcoder. The results of the evaluation are encouraging and demonstrate that our model achieves and surpasses the state of the art in the feature matching on low-textured objects.

Museums all over the world store a large variety of digitized paintings and other works of art with significant historical value. Over time, these works of art deteriorate, making them lose their original splendour. For paintings, cracks and paint losses are the most prominent types of deterioration, mainly caused by environmental factors, such as fluctuations in temperature or humidity, improper storage conditions and even physical impacts. We propose a neural network architecture for the detection of crack patterns in paintings, using visual acquisitions from different modalities. The proposed architecture is composed of two neural network streams, one is a fully connected neural network while the other consists of a multiscale convolutional neural network. The convolutional neural network plays a leading role in the crack classification task, while the fully connected neural network plays an auxiliary role. To reduce the overall computational complexity of the proposed method, we use morphological filtering as a pre-processing step to safely exclude areas of the image that do not contain cracks and do not need further processing. We validate the proposed method on a multimodal visual dataset from the Ghent Altarpiece, a world famous polyptych by the Van Eyck brothers. The results show an encouraging performance of the proposed approach compared to traditional machine learning methods and the state-of-the-art Bayesian Conditional Tensor Factorization (BCTF) method for crack detection.

The VERONA (Van Eyck Research in OpeN Access) research project enables open access to high resolution scientific image data of the oeuvre of the Flemish artist Jan van Eyck (ca. 1390-1441). The project is carried out by the Royal Institute for Cultural Heritage (KIK-IRPA, Brussels) in collaboration with the museums that have Van Eyck artworks in their collections. In order to accomplish and stimulate new research on the production of Jan van Eyck and his workshop, the image data acquired during the project is made publicly available via the website 'Closer to Van Eyck', as a supplement to the section dedicated to the conservation treatment of the Ghent Altarpiece.

The Belgian Art Links and Tools (BALaT) (http://balat.kikirpa.be/) is the continuously expanding online documentary platform of the Royal Institute for Cultural Heritage (KIK-IRPA), Brussels (Belgium). BALaT contains over 750 000 images of KIK-IRPA’s unique collection of photo negatives on the cultural heritage of Belgium, but also the library catalogue, PDF’s of articles form KIK-IRPA’s Bulletin and other publications, an extensive persons and institutions authority list, and several specialized thematic websites. All these are interlinked to give the user easy access to freely available information on the Belgian cultural heritage. The last years KIK-IRPA has been working on a detailed and inclusive data management plan. Through this data management plan, a new project HESCIDA (Heritage Science Data Archive) will upgrade BALaT to BALaT+, enabling access to searchable registries of KIK-IRPA datasets and data interoperability. BALaT+ will be a building block of DIGILAB, one of the future pillars of the European Research Infrastructure for Heritage Science (E-RIHS), which will provide online access to scientific data concerning tangible heritage, following the FAIR-principles (Findable-Accessible-Interoperable-Reusable). It includes and enables access to searchable registries of specialized digital resources (datasets, reference collections, thesauri, ontologies, etc.); supports data interoperability through the creation of shared knowledge organization systems and provides tools to process them according to researchers’ needs and research questions. Such data will primarily consist out of measurement results and scientific information (texts, images, 3D models and more). Special attention will be given to the implementation of the IIIF protocol to enhance our research environment with not only access to KIK-IRPA’s high-resolution images of cultural heritage, but also making it possible to include images from other institutions that are IIIF compliant.

Physical distortions (such as thorn-offs and scratches) are commonly seen in historical documents. Their presence disturbs downstream processes such as optical character recognition (OCR) and layout analysis, which leads to reduced productivity in automatic document information retrieval. A proper characterization of such physical noise is an important step in the development of historical document denoising methods. In this paper, we tackle noise characterization with Bayesian labeling, where noise and text pixels are characterized in terms of likelihood densities. We employ in particular two different significance measures, which are formulated using pointwise and cone-of-influence (COI) approximation of local Lipschitz regularity in the wavelet domain. We evaluate the effectiveness of the proposed noise characterization using a binary noise versus text classification model, where we show that a naive binary classifier using average point ratio (APR) or average cone ratio (ACR) distribution densities leads to effective classification of noise and text pixels with encouraging overall success rates. This encourages future work on the development of Bayesian frameworks for the recognition of physical distortions in historical documents.

Multi-light, single-camera imaging techniques like Reflectance Transformation Imaging (RTI, including PTM, HSH, and PCA-RBF) and the Portable Light Dome (PLD) have been used by cultural heritage scholars and collection curators extensively because of the extra interactive visual information that can be revealed on artefacts when compared to standard digital photography. Besides a virtual relighting of the scanned object, these techniques offer filters to accentuate different aspects of the studied surface. The main focus of RTI, developed at HP, CHI and among others elaborated by ISTI CNR, is aimed at photo-realistic virtual relighting. PLD, developed at KU Leuven, on the other hand, is aimed at extracting surface properties such as surface color (albedo), surface gradient (normals), 3D surface features (height profiles) and reflectance distribution (reflectance maps). PLD and RTI both produce interactive pixel based file formats, which are dissimilar, resulting in incompatible datasets. The pixel+ project (Art and History Museum, KU Leuven, and KBR, Belspo BRAIN-be funded) aims to merge both technologies into one web-based consultation platform, allowing existing PLD and RTI datasets to be viewed in one web environment with their respective viewing filters as well as to illuminate the virtual model. Moreover, as both methods are alike in terms of required input and processed output, pixel+ focuses on other types of integration, resulting in new viewing modes for processed data as well as a novel reprocessing pipeline for existing source images. In addition, for sustainable and flexible web consultation a new open format, based on glTF, is suggested and a first elaboration is presented.

In the restoration process of classical paintings, one of the tasks is to map paint loss for documentation and analysing purposes. Because this is such a sizable and tedious job automatic techniques are highly on demand. The currently available tools allow only rough mapping of the paint loss areas while still requiring considerable manual work. We develop here a learning method for paint loss detection that makes use of multimodal image acquisitions and we apply it within the current restoration of the Ghent Altarpiece. Our neural network architecture is inspired by a multiscale convolutional neural network known as U-Net. In our proposed model, the downsampling of the pooling layers is omitted to enforce translation invariance and the convolutional layers are replaced with dilated convolutions. The dilated convolutions lead to denser computations and improved classification accuracy. Moreover, the proposed method is designed such to make use of multimodal data, which are nowadays routinely acquired during the restoration of master paintings, and which allow more accurate detection of features of interest, including paint losses. Our focus is on developing a robust approach with minimal user-interference. Adequate transfer learning is here crucial in order to extend the applicability of pre-trained models to the paintings that were not included in the training set, with only modest additional re-training. We introduce a pre-training strategy based on a multimodal, convolutional autoencoder and we fine-tune the model when applying it to other paintings. We evaluate the results by comparing the detected paint loss maps to manual expert annotations and also by running virtual inpainting based on the detected paint losses and comparing the virtually inpainted results with the actual physical restorations. The results indicate clearly the efficacy of the proposed method and its potential to assist in the art conservation and restoration processes.

In this paper, a color holographic imaging solution using tabletop holographic display and synthetic aperture Fourier capture system is discussed. As a display, we utilize a tabletop prototype, which is based on temporal multiplexing method and two parabolic mirrors from our previous work. In order to provide the same perspective effect as looking at an object on a table, the wavefield of the tabletop display propagates from the hologram plane to a viewing window at 45° angle. A color capture system is implemented by optical path modification of synthetic aperture Fourier capture system using two tilting mirrors to record the same perspective view of the object as the display. The numerical processing path for three channel color hologram solution is presented including compensation of lateral RGB dimensions, propagation with sampling compensation, and binarization. To evaluate the proposed method, numerical and optical reconstructions are presented in the experimental results.

Spatial light modulators (SLMs) are the most common used digital elements for electro-holographic display, which allow presenting the computer-generated holograms (CGHs) without photosensitive medium and interference for recording. Most of the available SLMs can only modulate either amplitude or phase, which leaves complex amplitude modulation as a challenging problem to be solved. Since the method of generating holograms based on the decomposition of a complex value into two phase values was proposed in 1970, the encoding for a complex field is realized. The double-phase holograms (DPHs) can synthesize the two phase values in to a single hologram after sampling. DPHs are thus able to recover the complex object field approximately but sacrifice the accuracy and intensity of high spatial frequency components. To reconstruct the complete and accurate complex field, a high quality complex modulating method is proposed. The proposed method utilizes the interference of two phase-only wave-fronts modulated by two SLMs respectively. The prospective complex field is formed after the coherent superposition and sufficient long-distance diffractive propagation. The complex modulation and reconstruction based on phase interference is equivalent to the capacity of the amplitude and phase information for the complex object wave-front. This method is also free of speckle noise brought out by random phase encoding and the twin images reconstructed by the amplitude only holograms.

This paper introduces full bandwidth Coarse Integral Holographic Display architecture with spatial tiling for scalability. The new method uses a large resonant scanner in conjunction with spatial tiling of spatial light modulators (SLMs) to create doubling the hologram’s horizontal size and utilize the full bandwidth of the SLMs.

Novel research applications to holographic image projection using random lasers endow superior functionalities over reduced coherent artifacts and high-quality speckle free full-field image projections. Using conventional coherent light sources; leads to speckle or corrupt image formation and lacks in full-field holographic image projection. We quantitatively demonstrate solvent engineered low-cost efficient fabrication process to synthesize high quality and durable perovskite – polystyrene (wt. %) based random lasers with robust optical performances to obtain high-resolution speckle free computer-generated holographic (CGH) image projections. Further optical characterizations of perovskite – polystyrene (wt. %) based random lasers illustrate the addition of polystyrene results in wavy surface texture with dominant multiple scattering effects; that leads to a low threshold lasing operation. Besides this, we have also analyzed photon lifetime tests for different random lasers, excitation area dependent disorder strength induced threshold behavior, far-field analysis. Such random lasers are deemed to be greater potential applications to spectroscopy, optical coherence tomography, colored image formation, and image processing, etc.

Optically addressed spatial light modulator (OASLM) has attracted an increasing amount of attention due to its enormous potential in digital holography. Solution-processed ZnO nanoparticle (NP) is particularly promising in OASLM as a photosensor because ZnO NP-based OASLM can provide high resolution, large panel size, and economical production cost. However, it is still challenged by slow switching speed due to a long fall time (off time) and it takes tens of seconds for ZnO NP-based OASLM to switch off, which is limiting its application in dynamic digital holography. In this paper, the optimization in the fall time of ZnO NP-based OASLM is presented. The reduction in fall time was achieved by thermal annealing of ZnO NP, which decreases the effect of charge carrier trapping. Moreover, the fall time was further shortened by introducing a SiO₂ layer between ZnO NP and liquid crystal layer so that the charge trapping at the interface is weakened. The shortest fall time of 0.99s was achieved in reflective OASLM, incorporating PEDOT:PSS-ZnO NP heterojunction. The heterojunction was also characterized in terms of IV characteristics and its barrier height was measured by CV measurement. The reduction of fall time discussed in this paper is valuable and it makes ZnO NP-based OASLM attractive for dynamic holographic displays.

In the last two decades, Digital holography (DH) has been established as one of the most effective coherent imaging technologies and it has been widely used in different applications, as for example in 3D imaging and display, optical metrology and microscopy. However, due to the coherent nature of light sources, DH have to managing the issue of speckle noise, which often introduce remarkable degradation in holographic reconstructions. Speckle noise can be fixed in two main ways, i.e. reducing the coherence of light source (e.g. using Light Emitting Diodes) or implementing high performance de-speckling methods. This paper focuses on the second way by considering the recent advances in holographic image denoising. Currently, a universal denoising strategy for completely suppressing holographic noise has not yet been established. Therefore, the removal of speckle noise from holographic images represents a bottleneck for the optics and photonics scientific community. Here, the best-performing noise reduction approaches, that have been proposed so far, are presented, by paying particular attention to how such processing can affect highly demanded DH applications. Among them, hybrid solutions based on engineered holographic recording and suitable numerical processing are discussed and compared each other.

Fourier lensless holograms of large size and large viewing angle are investigated. The investigated hologram is non paraxial and has large size in horizontal direction and small in the vertical one. This paper focuses on the problem of efficiency and quality of numerical reconstruction of such large holograms. Three wavefront propagation methods are analyzed for this task: (i) Fresnel propagation, (ii) classical Angular Spectrum and (iii) Angular Spectrum with optimized zero padding. All the numerical approaches are investigated theoretically and experimentally for its numerical efficiency and reconstruction errors. Experimental verification is provided.

One hologram calculation method is the random phase-free method. When this method is used for an amplitude hologram, a reproduced image with a high-quality image can be obtained. However, when the random phase-free method is used for the kinoform, the reproduced image is degraded. In this study, we applied the kinoform encoding method proposed by Li to the random phase-free method to improve the reproduced image. The effectiveness of the proposed method was compared with that of the conventional method via simulations and optical experiments. Additionally, the parameters were optimized by the simulations, and the effectiveness was verified by numerical experiments.

We can record digitally-designed information of three-dimensional (3D) objects or optical elements on a holographic photosensitive material by using wavefront printing technology. But the hologram data generated from the digitally-designed information are very huge and there are often the occurrences of the unnecessary bidirectional communications. To solve this problem, we studied on a special-purpose computer for wavefront printing technology. This technique consists of generating the light-ray information from digitally-designed information of 3D objects, converting the light-ray information to the wavefront information and generating the hologram data locally from the wavefront information in interaction. In this paper, we designed the emulator of the special-purpose computer for wavefront printing technology and obtained the amount of information (the number of bits) required for the circuit by comparing the 3D images reconstructed from the holograms generated by the emulator. As a result, the amount of information of the wavefront information converted from the light-ray information most affected the quality of the 3D images reconstructed from the holograms generated by the emulator and we can design the emulator that can reduce the noise component from those 3D images. In the future, we will design the special-purpose computer for wavefront printing technology by using hardware description language and implement that special-purpose computer on a programmable logic device such as a field programmable gate array.

Phase-shifting interferometry is a highly accurate technique for obtaining phase distribution from the recorded fringe patterns. Generally, phase-shifting interferometry requires recording several fringe patterns with varying phase shifts experimentally and during the acquisition, the object must be stable. Also, the atmospheric turbulence and mechanical conditions should also remain constant during this time. These requirements limit the use of these phase-shifting interferometric techniques in dynamic event studies. In the present work, we introduce Riesz transformed based digital four-step phase-shifting interferometer to obtain phase distribution from a single recorded fringe pattern. All the experimental phase-shifting setups necessary to realize the phase-shifting are removed. The idea is based on the recording of a single fringe pattern, and computes its Riesz transform at first, second and third-orders. The obtained Riesz transform components are combined to generate three π/2 phase-shifted fringe patterns, and then, the phase distribution is obtained from these phase-shifted fringe patterns. The performance of this method is demonstrated first by using numerical simulation and the quantitative appraisal is given by using image quality index. Further, we apply this technique on a real fringe pattern recorded in digital speckle pattern interferometry (DSPI). The obtained results reveal that our method provides a simple and accurate solution for phase evaluation, therefore, makes it suitable for real-time measurements.

Nowadays, image and video are the data types that consume most of the resources of modern communication channels, both in fixed and wireless networks. Thus, it is vital to compress visual data as much as possible, while maintaining some target quality level, to enable efficient storage and transmission. Deep learning (DL) image coding solutions, typically using an auto-encoder architecture, promise significant improvements in compression efficiency. These methods adopt a novel coding approach where the encoder-decoder architecture is mostly based on neural networks, notably with analysis and synthesis transforms learned from a large amount of training data and an appropriate loss function. There are limited amount of works targeting the subjective evaluation of DL learning-based image coding solutions compression performance. Since learning-based image codecs use complex and highly non-linear generative models, very different artifacts are present in the decoded images, when compared to conventional artifacts such as blockiness, blurring and ringing distortions typical of traditional DCT block-based and wavelet image coding. In this context, the main objective of this paper is to review, characterize and evaluate some of the most relevant learning-based image coding solutions in the literature. Regarding the subjective quality evaluation, the assessment tests were conducted during the 84th JPEG meeting in Brussels, Belgium, by a mix of experts and naive observers. These subjective tests evaluated the performance of five state-of-the-art learning-based image coding solutions against four conventional, standard image coding (HEVC, WebP, JPEG 2000 and JPEG), applied to eight natural images, at four different coding bitrates. The experimental results obtained show that the subjective quality obtained with the selected learning-based image coding solution are competitive with conventional codecs. Moreover, a thorough inspection on the visual results has revealed some of the typical artifacts encountered in the learning -based image coding.

JPEG Systems aims to lay the groundwork to build uniform applications and extensions on top of the various JPEG coding standards. In 1992 JPEG released the first JPEG standard which is still the most commonly used image file format to date. However, over the years, JPEG developed several other standards that became popular in specific domains. JPEG 2000 for example is widely adopted in medical imaging, digital cinema and geographical information systems. The original JPEG file format supported extensions by using app marker segments. These app marker segments allow to embed chunks of data in an image file. The content will be ignored if the it is not recognized by the decoder. Hence, it allows to build extensions while retaining backwards compatibility with legacy decoders. JPEG 2000 and several other newer formats on the other hand use box based formats. Boxes are containers of data with a specific type that allow more flexibility with respect to length and structure compared to marker segments. Again, if the decoder doesn't recognize the particular box type the content will be ignored. JPEG XT defines extensions to the legacy JPEG file format such as transparency, HDR support and a lossless mode. To this end, JPEG XT introduced a file format which allows to embed boxes in app marker segments. Using this principle, new extensions can be uniformly defined across the various JPEG file formats, i.e. by defining boxes. Therefore, JPEG Systems defines tools and extensions by the definition of boxes. The JPEG Universal Metadata Box Format (JUMBF) allows to embed any type of metadata in any type of JPEG image file and supports the possibility to reference this data from within or outside this image. Using this structure, JPEG Privacy and Security, 360 and JLINK define mechanisms to support privacy and security tools, 360 images and linked images respectively. This paper gives an overview of the JPEG Systems architecture and illustrates its working by introducing JPEG 360 and the recently initiated activity JLINK.

Privacy and security, copyright violations and fake news are emerging challenges in digital media. Social media and data leaks increase risk of user privacy. Creative media particularly images are often susceptible to copyright violations which poses a serious problem to the industry. On the other hand, doctored images using photo editing tools and computer generated images may give a false impression of reality and add to the problem of fake news. These problems demand solutions to protect images and associated metadata as well as methods that can proof the integrity of digital media. For these reasons, the JPEG standardization committee has been working on a new Privacy and Security standard that provides solutions to support privacy and security focused workflows. The standard defines tools to support protection and integrity across the wide range of JPEG image standards. Related to image integrity, blockchain technology provides a solution for creating tamper proof distributed ledgers. However, adopting blockchain technology for digital image integrity poses several challenges at the technology level as well as at the level of privacy legislation. In addition, if blockchain technology is adopted to support media applications, it needs to be closely integrated with a widely adopted standard to ensure broad interoperability. Therefore, the JPEG committee initiated an activity to explore standardization needs related to media blockchain and distributed ledger technologies (DLT). This paper explains the scope and implementation of the JPEG Privacy and Security standard and presents the current state of the exploration on standardization needs related to media blockchain applications.

High Throughput JPEG 2000 (HTJ2K, JPEG 2000 Part 15) is a new part of the JPEG 2000 family of standards. HTJ2K provides an alternate block coding algorithm that can be used in place of the EBCOT algorithm, which is used in the JPEG 2000 Part 1. Similar to JPEG and HEVC-Intra mode, a desired option for the proposed HTJ2K (which is not available in the previous versions of the JPEG 2000) is to have a simplified Quality Factor (QF) to steer the compression process. This way the user can modify the compression level by simply increasing or decreasing the value of the QF without knowing the exact underlying compression ratio or the utilized bitdepth. In this paper, we report the results of a subjective experiment, which was designed and implemented to test a proposed QF function for the new HTJ2K encoder, according to the Q factors of the JPEG encoder. A new test methodology for this purpose is utilized and a new formulation of the Q-factor is derived for HTJ2K based on the experimental results. The statistical analysis of the experimental results beside the procedure for deriving the QF range are provided. The derived QF range in this experiment is planned to be integrated in the implementations of the High Throughput JPEG 2000.

JPEG XL is a practical, royalty-free codec for scalable web distribution and efficient compression of high-quality photographs. It also includes previews, progressiveness, animation, transparency, high dynamic range, wide color gamut, and high bit depth. Users expect faithful reproductions of ever higher-resolution images. Experiments performed during standardization have shown the feasibility of economical storage without perceptible quality loss, lossless transcoding of existing JPEG, and fast software encoders and decoders. We analyse the results of subjective and objective evaluations versus existing codecs including HEVC and JPEG. New image codecs have to co-exist with their previous generations for several years. JPEG XL is unique in providing value for both existing JPEGs as well as new users. It includes coding tools to reduce the transmission and storage costs of JPEG by 22% while allowing byte-for-byte exact reconstruction of the original JPEG. Avoiding transcoding and additional artifacts helps to preserve our digital heritage. Applications require fast and low-power decoding. JPEG XL was designed to benefit from multicore and SIMD, and actually decodes faster than JPEG. We report the resulting speeds on ARM and x86 CPUs. To enable reproduction of these results, we open sourced the JPEG XL software in 2019.

A new image compression standard, known as JPEG Pleno, is currently under its final standardization stage and has been developed by the ISO/IEC JTC 1/SC 29/WG 1 standardization committee. This paper reports the current status of the final draft of JPEG Pleno Part 1, Part 2 and a detailed description of the conformance tests and the reference software under standardization as Part 3 and 4.

Digital holography is an imaging technique with great potential, that allows the acquisition and the reproduction of the entire wavefield of light. A big challenge in this field is the development of efficient coding tools, due to the different hologram data characteristics compared to standard multimedia content. This work presents an experimental comparison between two different coding techniques: the well-known hologram plane coding and the less-investigated object plane coding. In order to identify advantages and limitations of each method, but also to determine if the object plane coding could be an effective alternative to the more simple hologram plane coding, different types of holograms are employed, analyzing and discussing the obtained results.

Curved sensors are a disruptive technology which allow improvement of the image quality and substantial simplification of optical designs. Both bring to the reduction of the volume and mass of the instruments, while improving their performance. The field of curvature is an optical aberration, present in every single optical system, describing the fact that the light going through optics produces an image on a portion of a sphere. The impact of this aberration is seen mostly on the edges of an image by the presence of vignetting and distortion. Current solutions are applied to correct the images, but they imply a growing complexity of the instruments and the increase of their volume. Curved detectors correct this optical aberration directly at the focal surface. Our recent research has demonstrated that curved detectors allow to remove many lenses and release constraints in optical systems, simultaneously making them more compact without performance impairments. The CNRS-LAM has developed a nondestructive, high quality curving process. A full electro-optical characterization performed at LAM on prototypes demonstrates that there is no significant difference among curved and flat sensors. The study and achievement of the association of a wide-angle lens with a curved sensor have proven that it is possible to reduce by a factor of two the number of lenses in the optical system. Moreover, a significant improvement of the sharpness can be noticed, just like the fact that there is no more distortion or vignetting, all until the edges of the image. Curved sensors bring significant improvement of the image quality in trans-disciplinary applications. The team is currently developing new optical concept including a curved focal surface for projects going from bio medical application to space imaging. In this talk I will present our work and results in the frame of freeform CMOS development, optical concept for a wide field imager for an ESA Earth’s aurora monitoring mission and discuss the impact of curved sensors on imaging systems.

Video surveillance systems for the main production processes at nuclear power plants are an important research topic in our time. Reliability, quality, and radiation resistance – are among the main fundamental requirements for the created products. Inspection of many areas that are important for safe operation can be carried out only with the help of TV cameras based on the vidicon tube receiver, which form only a grayscale image. The ability to convert a grayscale image obtained with radiation-tolerant television cameras into a color image will significantly increase the reliability and information content of the control, which will allow an objective assessment of the condition of the equipment, and as a result will reduce the risk of emergency situations. The solution proposed in the work is a technique for converting grayscale image obtained with radiation-tolerant cameras based on vidicon tube receiver into color one. One of the main characteristics of the vidicon is its spectral characteristic, which is determined by the device and the properties of the photographic target.

Obtaining accurate and noise-free three-dimensional (3D) reconstructions from real world scenes has grown in importance in recent decades. In this paper, we propose a novel strategy for the reconstruction of a 3D point cloud of an object from a single 4D light field (LF) image based on the transformation of point-plane correspondences. Considering a 4D LF image as an input, we first estimate the depth map using point correspondences between sub-aperture images. We then apply histogram equalization and histogram stretching to enhance the separation between depth planes. The main aim of this step is to increase the distance between adjacent depth layers and to enhance the depth map. We then detect edge contours of the original image using fast canny edge detection, and combine linearly the result with that of the previous steps. Following this combination, by transforming the point-plane correspondence, we can obtain the 3D structure of the point cloud. The proposed method avoids feature extraction, segmentation and the extraction of occlusion masks required by other methods, and due to this, our method can reliably mitigate noise. We tested our method with synthetic and real world image databases. To verify the accuracy of our method, we compared our results with two different state-of-the-art algorithms. In this way, we used the LOD (Level of Detail) to compare the number of points needed to describe an object. The results showed that our method had the highest level of detail compared to other existing methods.

Automotive and robotic applications demand three-dimensional (3D) imaging systems, that are reliable, small and lowcost. A promising technology to satisfy these demands are LiDAR-systems based on the direct time-of-flight principle. To greatly reduce the costs and footprint of current LiDAR-systems new compelling concepts are emerging. These concepts mainly differ in the used illumination techniques of the field-of-view (FOV). Micro-electro-mechanical systems (MEMS)- scanners are replacing the bulkier and more expensive polygon mirrors, which were earlier used to deflect the laser beam and partially illuminate the FOV. In contrast to this approach the full illumination of the FOV with a single widened laser beam is used by the concept known as flash-illumination. The range performance of the direct time-of-flight principle is directly proportional to the emitted laser pulse power and is either limited by detector noise or background radiation. High-power near-infrared laser diodes are announced, which offer higher pulse peak power and shorter laser pulses. The high energy density of these laser sources can, however, be hazardous to the human eye. Therefore, the energy density of the laser pulse must be limited and classified according to the IEC 60825-1:2014. While this is particularly easy for Flash LiDAR-systems, a two-dimensional consideration is necessary to determine the accessible emission limits (AEL) for MEMS-based LiDARsystems, which takes the time dependent scanning-motion of the mirror into account. This work presents a framework to describe and calculate the differences between the calculations of the AEL for Flash- and MEMS-based LiDAR-systems. To achieve this the framework incorporates the calculation of the two-dimensional pulse distribution based on the time dependent behaviour of the MEMS-scanner for the most commonly used one- and two-dimensional scanning-patterns. To further estimate and compare the performance of the LiDAR-systems this work describes an automotive and robotic application use case scenario, which are used to compare the performance for various performance metrices.

The rise of interest in Building Information Modelling (BIM) during the last years has led to the introduction of complex data structures. Among all this information, a precise definition of structures in the form of point clouds is still an open research area. Consequently, an accurate, fast and easy to use point cloud acquisition system is needed. In this paper, we present a hardware-software prototype for automated indoor environments scanning through the acquisition of point clouds using multiple low-cost depth sensors. The acquisition and processing phase are described in detail, as well as the test environments used to assess the system functionality.

For long, the VR optics designer has been taking advantage of the fact that the eye is not optically perfect and striving towards creating “perfectly imperfect” VR optics. These cheaper and lighter lens systems could have high off-axis aberration, but with negligible impact on visual quality. A good way to test and validate these designs could be simulating the rendered image of the pixel array on the retina. Earlier attempts used the principle of virtual ray tracing, which is inherently a Monte Carlo method and therefore subjects to noise at sharp transition edges. Furthermore, this approach cannot render the Point Spread Function (PSF). We took the reverse approach of physical ray tracing. Rays are generated from each illuminating point and traced forward to the retina. After the tracing process, each light source point generates a number of ray-retina intersections. We propose a novel rasterization algorithm to render the radiometric image on the virtual retina. For the radiometric-correct determination of pixel value, we used the Intersection-over-Area metric. This metric is an indicator for how much the back-projected area of each retinal cell overlaps the triangle formed by the cone of light flux falling on the retina. The geometry of the overlapping was found with the Sutherland-Hodgman algorithm. All images of source points were then superposed to find the resulting image. With the proposed software, it is possible to simulate different VR-optics in combination with eye models. The most significant challenge of the physical ray tracing was the huge calculation workload. Our software overcame this by utilizing parallel GPU computing capability offered by the CUDA platform. For evaluation, we used the software with different schematic models of the eye, including those from Gullstrand, Le Grand, Koojiman and Navarro. We found that the former two, so-called paraxial, eye models predict the eye has higher sagittal resolution than tangential. On the other hand, the latter two (so-called finite eye models) predict the opposite. With this simulation software, optical designers can validate their VR-optics-design and gain more insights about the image formation process. This tool can be used as a virtual optical test bench for the simulation of the imaging result before building a costly prototype.

This paper proposed a multi-channel spectral coding method for the coded aperture tunable filter spectral imager where the liquid crystal tunable filter is switched to encompass several selective spectral channels into one snapshot. The spectral coding and spatial coding could be designed concurrently.

Fluorescence-lifetime spectroscopy has been recently used in different applications in rapid medical diagnoses for different diseases, e.g. cancer, oral carcinoma, actinic cheilitis and tissue diagnosis. In addition, it has various applications in biology, chemistry analysis, pharmaceutical applications and physics. The existing fluorescence spectrometer systems are bulky, contain complex optical setups, large and use either monochromators with movable parts, hence less mechanically stable, or use non-tunable bandpass filters. Therefore, the user has to adjust the filters while changing the fluorophore. In this work, we propose a spectrometer setup that can be tuned for different emission wavelength using spatial light modulators (SLM). SLMs use highly stable MEMS micromirrors in their architecture, hence, the system has no large moving parts and is mechanically stable. In addition, the system is suitable for more fluorophores and thus widens the application potentials of the system. Digital micromirror devices (DMD) are widely used in optical projection systems. However, some recent developments were made to use them in conventional optical spectroscopy which enabled using single point detectors to collect the spectrum instead of using linear or area camera sensor arrays. This is expected to have a big effect on the cost of the spectrometer. Using similar techniques, in this work, a light source, e.g. a supercontinuum laser was used along with dichroic filters to fix the excitation wavelength at a certain value. After that, the selected wavelength(s) pass through a beam splitter towards the sample. The beam splitter is installed to separate the emission and excitation light. The fluorescence emission is reflected from the beam splitter and is focused on a diffraction grating. The different wavelengths are projected to the DMD. The used DMD has a VGA resolution with a mirror pitch of 5.4µm. Each micromirror could be controlled individually into two states, on and off at ±17 degrees. The setup ensured that different wavelengths of light always fall on different mirror(s) and hence the desired wavelength(s) could be selected at once. A high-speed-gated single pixel CMOS detector with a large diameter was used to capture the reflected light with the desired wavelengths from the DMD. The system scans the mirrors and hence the wavelengths to readout the spectrum. In addition, the system measures the fluorescence lifetime of the used fluorophore at the emission wavelength. The proposed setup is compact and less complicated than conventional systems. Furthermore, the spectrometer, apart from the light source, has a great potential to be miniaturized. Thus, it can be used as hand-held system. The proposed system was characterized and the proof of concept of the setup was verified. Due to the small pitch of the DMD and the large number of mirrors, it was found that it has a high resolution especially in the VIS spectral range. The system is considered as semi-tunable as it can be tuned for different emission wavelengths and thus, can be used with different fluorophores. However, the light source along with the filters should be changed for different excitation wavelengths. or pulsed laser diodes can be used.

Adaptive coding is a recently emerging strategy for designing coded apertures using a-priori information of target layout to improve the reconstruction accuracy of compressed information. Based on the compressive sensing imaging system, this paper proposes a method for designing adaptive coded apertures by using the low-resolution a-prior information of every spectral band of the scene. This method also enables coded apertures containing both positive and negative coding information can be implemented in a compressive spectral imaging system that uses DMD or similar device as a coding device. Compared with other designs that utilized various kinds of a-prior information, this solution does not require additional equipment and is in real-time. The simulation results show the proposed real-time adaptive coded apertures method can attain better reconstruction performance than that of a random coding method at relative high compression ratio.

Non-invasive skin cancer diagnostic methods develop rapidly thanks to Deep Learning and Convolutional Neural Networks (CNN). Currently, two types of diagnostics are popular: (a) using single image taken under white illumination and (b) using multiple images taken in narrow spectral bands. The first method is easier to implement, but it is limited in accuracy. The second method is more sensitive, because it is possible to use illumination considering the absorption bands of the skin chromophores and the optical properties of the skin. Currently CNN use a single white light image, due to the availability of large datasets with lesion images. Since CNN processing and analysis requires a large image database, only mathematical models have been used for multispectral diagnostics so far. Several scientific groups have created unique CNN, but the possibility of sharing pre-trained CNN models is limited due to the diversity of spectral bands used for skin lesion imaging. Current CNN models require image sets where each skin lesion has the same number of specific spectral bands. Therefore, researchers are unable to share their trained CNN models and each team uses a limited amount of skin lesions for CNN training. The paper proposes multi-input CNN architecture with a special encoding layer that allows using images in arbitrary spectral bands. That will allow sharing pre-trained models with other researchers to fine tune the model using additional wavebands. Additionally, the proposed model can adopt images taken under white illumination. As a result, it will be able to increase current melanoma detection accuracy

With the development of sensor technologies, imaging technology is developing more rapidly. What followed was the widespread use of image processing technology in many kinds of applications. For instance, image processing technology has been widely used in video surveillance, medical diagnosis, remote sensing detection and object tracking. As a sub-field of image processing technology, image fusion is the one of most studied technology. The aim of image fusion is to acquire an integrated image that contains more information. This integrated image is more conductive for a human or a machine to understand and mine the information contained in the image. In all kinds of image fusion, infrared (IR) and visible (VIS) image fusion is one of the most valuable multisource image fusion. When imaging the same scene using both IR and VIS imaging system, more information can be obtained, but more redundant information is generated. The IR sensor acquires the thermal radiation information of the object in a scene, so the object can also be detected when the lighting conditions are poor. The image acquired by VIS light sensors has more spectral information, clearer texture details, and higher spatial resolution. Thus, the scene can be described more completely by integrating the IR and VIS images into one image. Meanwhile, the scene can be readily understood by observers, and the information of the scene can be easily perceived. In this paper, an effective IR and VIS image fusion via non-subsampled shearlet transform (NSST) and pulse-coupled neural network (PCNN) in multi-scale morphological gradient (MSMG) domain is proposed. First, low frequency sub-image and high frequency sub-images are obtained through NSST. Then, the low frequency sub-image and high frequency sub-images are fused via a MSMG domain PCNN (MSMG-PCNN) strategy. Finally, the fused image is reconstructed by inverse NSST. Experimental results demonstrate that the proposed MSMG-PCNN-NSST algorithm performs effectively in most cases by qualitative and quantitative evaluation.

A technique for obtaining subpixel resolution when tracking through cross-correlation consists of interpolating the obtained function and then refine the peak location. Although the technique provides accurate location results, the peak is always biased towards the closest integer coordinate. This effect is known as peak-locking error and is a major limit to the experimental accuracy of this calculation technique. This error may be different depending on the algorithm used to fit and interpolate the correlation peak but no systematic analysis was found in the literature. In our study we explore the three most common interpolation methods: thin-plate splines, second-order polynomial fit and Gaussian fit together with the influence of the extent of local interpolation area around the peak. Additionally, we have checked the influence of the image blurring on the results, since it is reported as one effective method to reduce the peak-locking error. Finally, the optimal adjustment found is the Gaussian fit with no blur and a neighborhood around the correlation peak of 11x11 pixels.

Generally, classic longslit spectrometers consist of an entrance slit, a collimating lens, a disperser, and a projection lens. High spectral resolution is the main quality function of spectral devices; however, there is also a possibility to achieve high spatial resolution along the slit. Many scientific tasks require high spatial resolution along with high spectral one in astronomical instrumentation. For that, the integral field spectroscopy can be applied. The possibilities of an integral field unit (IFU) optical system synthesis are investigated in this work. The IFU optical system is supposed to be connected to a Solar Telescope-Coronagraph (the main mirror diameter is equal to 3 meters), which operates in a wide spectral range (0,39 - 1,60 microns). Application of integral field units for increasing the spatial resolution of spectral astronomical devices is formulated. The integral field spectroscopy technique description is given as well as the principal concept is shown. Classification of the IFU optical systems (lenslet-, fiber-, mirror-based) is provided. For systems, operating in a wide spectral range, the allreflective design is considered. The system design where the input telescope field (6′′ x 12′′) is divided into 64 slices (0,09′′ x 12′′ each) is proposed. The research results conclude in a computer model of the optical system of the integral field unit with the paraxial optics approach. The proposed solutions may be utilized in the design of optical slicing systems for various applications.

We present a compact dual-view endoscope imaging system with a field of view (FOV) of ±80° and F/# of 3.4. The endoscope consists of two optical configurations for increasing FOV within the volume constraint. The front view configuration is a fisheye lens with a FOV of ±55°, and the side view configuration is a panoramic annular lens that covers the remaining FOV. The two configurations are combined by a hybrid lens that consists of center refractive portion and side catadioptric portion. Both the front and rear surfaces of the hybrid lens are aspherized with the use of annularly stitched Q-type aspheres. Thus, a compact endoscope is successfully implemented with fewer lenses, with a total length of 11.5 mm and a maximum diameter of 5.5 mm. The modulation transfer function at 167 lp/mm is above 0.4 over the entire FOV. The relative illumination is more than 0.65 and the optical distortion is within 10%. Moreover, the near telecentric condition is fulfilled and supports constant magnification focusing.

Wave-optics and ray-field tracing methods for the simulation of micro-lens arrays (MLAs) taking into account the coherence and polarization effects of light source, randomization of microlens array parameters are implemented. The influence of the parameters of radiation source (wavelength, wavefront curvature, beam radius, coherence radius, etc.) and micro-lens array (periodic or random, aspect ratio, pitch size, refractive index, array surface shape and profile (convex, concave), etc.) on the output parameters (intensity distribution, radiation pattern, optical efficiency) of a diffracted beam is investigated. Numerical simulations of the intensity distributions and spreading angle of a diffracted beam have been carried out. The new ray-field approach based on the coherent states representation is developed for calculation of the optical efficiency of the microlens arrays. Such wave beams can be tracked simply through the arbitrary curved surfaces. User Friendly Interface is developed for introducing initial parameters of light source and MLA array and for graphical and data outputs of simulated results.

This article describes eye tracking technology as a prospective application for AR systems, opening opportunities for dynamical focus, attention mapping and the interplay between a user and a device. Described eye tracking system is an integration of a microcontroller and a camera module. The software uses algorithms for detection a pupil, determination of its position and calculation of the gaze vector for each eye. Gaze tracking is a basis for such technologies as dynamical focus and attention maps, which use the sight vectors as initial data. AR system with basic tracking functionality is developed on the base of multifunctional microcontroller Raspberry Pi which makes project available for a wide community of students and developers by minimal cost.

Object tracking with subpixel accuracy often needs targets with special shapes, which include random dot patterns, circular objects or targets with irregular contours. Unfortunately, in majority of real applications no such objects are found or cannot be attached to the region of interest and non-optimal objects must be used. Among those, we will analyze here the performance of rectangular objects that may be common in natural or artificial targets. Object tracking has been performed through the two most common methods that appear in the literature: centroid calculation and cross-correlation with peak interpolation. Numerical simulations show that tracking results for such objects are highly dependent of the object orientation with respect to the direction of movement. This is due to the interference of the object borders and the sensor and how is the subpixel information obtained. Best results are obtained for object orientations from 5 to 30 deg with respect to the normal to the displacement directions. Experimental results confirm the simulations and allow us to establish that, although object alignment provides a better image on the sensor with sharper borders, this situation is not desirable for accurate subpixel tracking, with accuracies varying from accuracies of 0.08 px (RMS error) with aligned objects to 0.02px with a misaligned target.

The Strasbourg School of Architecture (ENSAS) developed a daylight simulator applied to architectural scale models. It reproduces both direct sunlight and diffusing light from the sky. Because LED technology offers a possible control of each source, it is no longer necessary to build the sky with a dome shape. Therefore we chose a cubic shape. RGB LED sources are fixed on a white panel behind a lambertian diffuser. To reach a uniform lighting on the technical platform, the system has to be calibrated. The objective is to define the correction table to be applied to the artificial sky. After calibration in such a sky, the luminous flux of each source will have the same contribution to the centre of the experimental platform. Initial results have shown that the application of a theoretical correction function does not provide uniform illuminance. These first counter-intuitive results led to investigate several configurations which showed that the inter-reflection between the walls represent about 20% of the illuminance measured inside the simulator. Finally, because of these inter-reflections, the intensity of the LEDs in the corners should be reduced by around 10%. To respect a lighting ratio of 6 between the sun and the sky, the intensity of the artificial sun has been adapted.

Some materials undergo an hygric expansion when they are soaked. In porous rocks, this effect is enhanced by the pore space that allows the water to reach every part of its volume and to hydrate the most of their swelling parts. This enlargement has negative structural consequences in the vicinity since adjacent elements will support some compressions or displacements. Recently image-based methods have arisen in this field due to their advantages versus traditional methods. Among all image processing methods, digital image correlation (DIC) is one of the most used in all areas. In this work, we propose a new methodology based on DIC for the calculation of the hygric expansion of materials. We use porous sandstone, with dimensions 14x14x30 mm to measure its hygric swelling using an industrial digital camera and a telecentric objective. We took one image every 5 minutes to characterize the whole swelling process. Due the large magnification, the whole 14 mm length of one contour was not in the image and therefore we lost the image scale reference. To solve this, a 1951 USAF test was used to calibrate the imagen. The telecentric objective and a narrow deep of field allowed to have the specimen surface exactly on the same plane that the USAF test was during the calibration. The image was pointed to one corner of the specimen, to obtain information not only of its vertical displacement due to its expansion but also of its horizontal movement. Preliminary results show that the proposed methodology provides reliable information of the hygric swelling using a non-contact methodology, with an accuracy of 1 micron.

The study of the blink can provide a lot of information regarding ocular surface health, cognitive status, psychological health, human neurological disorders, etc. Blinking can be classified as spontaneous, reflex and voluntary. Between the reflex blinks, direct and consensual reflex can be categorized, being direct that from the eye that is being stimulated and by consensual the other one. A system to evaluate of spontaneous and reflex to light blinking is proposed using a led flash and a high-speed camera that records the blinking of each eye. The LED flash only illuminated one eye and alternated arbitrarily. The light intensity diffused by the eye is almost constant when the eyelid is open but changes when the eyelid closes until reaching a maximum peak, and so does the intensity that was registered by the camera. This variation is directly related to the variation of the eyelid position. Several parameters characterizing the blinks have been defined from the fitting of the blink curve to an exponentially modified Gaussian function. Sequences of 60 seconds of 9 subjects blinking were recorded. Results show that there are no significant differences between reflex and spontaneous blinking, however, significant differences between direct and consensual reflex blinking have been found.

The article describes the variants for using the relief phase diffraction gratings for a waveguide used as an optical input/output element for images in augmented reality systems. A characteristic feature of diffraction gratings with a surface relief, which, in this case, were recorded in the photoresist layer, is a relatively small relief depth (of the order of several tenths of a micrometer), while the diffraction efficiency of such gratings is quite high (up to 30%). The paper considers an implementation option for a waveguide with a common angular field of view close to 60°. The article presents the waveguide options used to output both monochrome images and display full-color images.

An experimental study of the color reproduction quality of RGB color components, white and some “composite” colors in the developed augmented reality device based on two waveguides with diffraction elements was carried out. The color gamut of the device is determined. Color characteristics were analyzed by color coordinates in the xy system, presented on the chromatic diagram, and are given in the work.

In operating rooms (OR), physicians have to work in strict compliance to asepsis rules to not endanger the health of patients. In the case of laparoscopic mini-invasive surgery, where the field of view of the surgeon is restricted, the use of computers is necessary to provide the missing information. Physicians must therefore interact with computers either by directly manipulating mouse and keyboard, by removing their gloves or by using a protection for the devices, or by voicecommanding an assistant to do so. However, in addition to be time-consuming, it may cause hygiene issues in the first case and a lack of precision in the second. The need to have better way of interactions with computers had led to important researches in that area during the last ten years, especially in Touchless Human Machine Interaction (THMI). Indeed, THMI, including gesture recognition, voice recognition and eye-tracking, has a promising future in the medical field, allowing surgeons to interact by themselves with devices, thereby avoiding error-prone process while complying to asepsis rules. In this context, the “Intelligent Touchless Glassless Human-Machine Interface” (ITG-HMI) project aims to provide a new tool for viewing and manipulating 3D objects. In this article, we present how this interface was implemented, through the detection and recognition of hand gestures using Deep Learning, the establishment of a graphical interface to display 3D models and the adaptation of gestures recognized in actions to achieve.

Image-based scene 3D reconstruction is one of the key tasks for many machine vision applications such as scene understanding, object pose estimation, autonomous navigation. A set of reliable and accurate methods for multi-view scene 3D reconstruction has been developed last decades. But a significant drawback of such 3D reconstruction technique is the need for acquiring a large number of images in the processed sequence to obtain an acceptable 3D scene representation. Recently modern convolutional neural network (CNN) models achieve the best quality for object recognition, image segmentation, image translation and some other challenging computer vision problems. The paper proposes a convolutional neural network architecture and a technique for training data preparation which provide a prediction of voxel model of a 3D scene with several objects. For CNN training and evaluation a special dataset was collected and annotated. It contains image sequences of several scenes and corresponding depth images and 3D models of these scenes. The image sequence serves as the primary data used for further scene 3D reconstruction by SfM technique. Structure from Motion processing results in surface 3D models of all objects in the scene and camera positions and orientation for every image in a sequence. Then surface 3D model is transformed into voxel 3D model and segmented into separate objects. Conditional generative adversarial network architecture was developed for 3D reconstruction by single image. Its generative part translates an input color image into an output voxel model. The discriminative part distinguishes the correct output (close to real voxel model) from false output (wrong output voxel model). Both parts are trained simultaneously on the prepared dataset. Evaluation on the testing part of the prepared dataset has demonstrated the ability of prediction 3D models of previously unobserved complex scenes containing several objects. The proposed neural network architecture provides high generalization ability and improved resolution of predicted voxel 3D models.

Ultrasound (US) has become one of the most common forms for medical imaging in clinical practice. It is a non-invasive and safe practice that allows obtaining images in real time. It is also a technology with important challenges such as low image quality and high variability (between manufacturers and institutions) [1]. This work aims to apply a fast and accurate deep learning architecture to detect and locate cerebellum in prenatal ultrasound images. Cerebellum biometry is used to estimate fetal age [2] and cerebellum segmentation could be applied to detect malformation [3]. YOLO (You Only Look Once) is a convolutional neural network (CNN) architecture for detection, classification and location of objects in images [4]. YOLO was innovative because it solved a regression problem to predict the location (coordinates and sizes) of bounding boxes and associated classes. We used 316 ultrasound scans of fetal brains and their respective cerebellar segmentations. From these, 78 images were randomly taken to be treated as test images and the rest were available to feed the trainings. Segmentation masks were converted to numerical descriptions of bounding boxes. To deal with small data set, transfer learning was done by initializing convolutional layers with weights pretrained on Imagenet [5]. We evaluated detection using F1 score and localization using average precision (AP) for 78 test images. Our best AP was 84.8% using 121 divisions or cells per image. Future work will focus on segmentation task assisted by localization.

Superpixel algorithms oversegment an image by grouping pixels with similar local features such as spatial position, gray level intensity, color, and texture. Superpixels provide visually significant regions and avoid a large number of redundant information to reduce dimensionality and complexity for subsequent image processing tasks. However, superpixel algorithms decrease performance in images with high-frequency contrast variations in regions of uniform texture. Moreover, most state-of-the-art methods use only basic pixel information -spatial and color-, getting superpixels with low regularity, boundary smoothness and adherence. The proposed algorithm adds texture information to the common superpixel representation. This information is obtained with the Hermite Transform, which extracts local texture features in terms of Gaussian derivatives. A local iterative clustering with adaptive feature weights generates superpixels preserving boundary adherence, smoothness, regularity, and compactness. A feature adjustment stage is applied to improve algorithm performance. We tested our algorithm on Berkeley Segmentation Dataset and evaluated it with standard superpixel metrics. We also demonstrate the usefulness and adaptability of our proposal in medical image application.

Due to circumstances beyond the presenter's control, an audio recording was not possible for this presentation.

We present optical design of wideband spectrometer for space-borne telescopes for science observation. Multi-channel configuration is adopted to cover wide wavelength region from visible to mid-infrared. Basic design of each channels are consisting of mirror optics and reflective spectral disperser. Low, medium, and high dispersion spectrometers are studied. Influence of defocusing is analyzed. Each channels are combined based on use of dichroic mirrors.

We demonstrate the performance of nonlocal means (NLM) and its related adaptive kernel-based methods for speckle denoising for the intensity and phase images acquired from the digital speckle pattern interferometric (DSPI) and digital holographic interferometric (DHI) techniques, respectively. The speckle denoise capabilities of NLM and its variant denoising methods such as NLM-average (NLM-av), NLM-local polynomial regression (NLM-PLR), and NLM-shape adaptive patches (NLM-SAP), and various NLM-reprojection schemes are implemented on simulated. Their performances are quantified on the basis of two metric criteria – peak signal-to-noise ratio (PSNR) and the image quality index (Q). The effectiveness of these denoising methods is compared with other existing speckle denoising methods. The obtained results suggest that these denoising methods have the ability to denoise speckles from the DSPI/DHI fringes and provide better visual and quantitative results.