Abstract Near‐infrared (NIR) meta‐optics often suffer from a limited field of view (FOV) and bandwidth in compact designs. While metasurfaces offer high design freedom in the lateral dimension, computational methods provide more flexibility for achieving precise control over the thickness dimension, offering a superior approach to reach the physical limits of the system. Here, a computational imaging method based on physics‐constrained embedded training, which integrates intrinsic physical constraints, including chromatic dispersion and fabrication tolerances, into the forward imaging simulation and neural network training, is proposed. This approach is demonstrated through the implementation of a NIR camera featuring a 6.8‐mm‐diameter monochromatic metalens, which, when coupled with a nonlinear activation‐free generative adversarial network, achieves broadband achromatic imaging with a FOV of 78° across the 800–1000 nm spectral range. The camera successfully captures real‐world scenes with high fidelity, enabling applications like vein detection.

This systematic mapping review and bibliometric analysis examines Scopus-indexed research on computer vision, image processing, and deep learning applied to wood and timber materials and products. A rule-based Scopus search (TITLE-ABS-KEY, 9 December 2025), combining wood and timber terms with imaging and computer vision terminology, followed by duplicate removal and structured exclusions, retained 1019 papers (1983–2026, early access) covering surface inspection, internal imaging, species identification, processing operations (log-yard/sawmill/panels), automation, dimensional metrology, and image-based property/structure characterization. The papers were classified into nine application categories and three methodological classes using improved rule-based classification with weighted scoring and exclusion rules. Paper output continues to accelerate, with 63.7% of papers published since 2016; Wood Surface Quality Control dominates (48.3%), followed by 3D and Internal Wood Imaging (13.6%), Wood Microstructure and Characterization (10.1%), and Wood Species and Origin Identification (10.6%). Methodologically, classical computer vision prevails (73.6%). Deep learning accounts for 26.4% of the corpus overall and 48.8% of papers from 2023–2026 (early access), while classical computer vision remains prevalent (70.1%) across most categories; the dataset totals 11,961 citations (mean: 11.74 per paper). Validation on 97 papers showed 80.41% accuracy for methodological classification and 70.1% for application categories. We quantitatively map method evolution across the nine categories, introducing a tailored taxonomy and tracking the shift from classical vision to deep learning at the category level. The remaining gaps include dimensional measurement automation, warp detection, sawing optimization, and benchmark datasets, with future directions emphasizing Vision Transformers, multi-modal sensing, edge computing, and explainable AI for certification.

Computational fluorescence microscopy constantly breaks through imaging performance through advanced optical modulation technologies; however, conventional theoretical modeling and experimental measurement approaches are challenging to meet the demand for accurate system characterization of diverse modulations. To this end, we propose a point spread function (PSF) decoupling method that is intrinsically compatible with the optimal demodulation in computational microscopic imaging modality. The critical core lies in designing a sample prior-based computational imaging strategy, in which a regular fluorescent sample instead of generally used sub-diffraction limited particles acts as a system modulator to demodulate the system response. PSF consequently can be computationally optimized through the strong support from the modulated sample prior, achieving accurate non-parametric system characterization and thereby avoiding the modeling difficulty and the low signal-to-noise ratio measurement errors of the system specificity. Experimental results across various biological tissues demonstrated and verified that the proposed PSF decoupling method enables excellent volumetric imaging comparable to confocal microscopy and multicolor, large depth-of-field imaging under aperture modulation. It provides a promising mechanism of system characterization and computational demodulation for high-contrast and high-resolution imaging of cellular and subcellular biological structures and life activities.

Computational polarization difference fingerprint imaging with stokes vector

Investigation of a Sink-Float plastic separator through Computational Fluid Dynamics and Particle Image Velocimetry.

Enhanced computational ghost imaging through dynamic turbid media via statistical averaging and deconvolution

High-dimensional computational imaging using spectral-polarization encoding and deep learning

Computational image analysis from the transverse plane of Drosophila embryos.

ABSTRACT We present a flexible and cost‐effective computational imaging framework for multichannel microscopy based on pupil plane multiplexing, enabling efficient spectral and polarization‐resolved imaging without bulky optical components. Conventional multichannel systems require sequential acquisition or hardware‐level multiplexing, often requiring complex optics or costly sensor modifications. Our method introduces a custom‐designed pupil division aperture at the microscope's Fourier plane, segmenting the pupil with distinct spectral or polarization filters to produce channel‐specific transfer functions. These encoded measurements are decomposed using a model‐based reconstruction algorithm that incorporates the channel‐dependent pupil functions as priors. We validate our approach through simulations across diverse pupil division geometries and experimentally demonstrate pupil plane polarization multiplexing using semicircular‐shaped linear polarizers and a single monochrome sensor. Integrated with Fourier ptychography, our method enables high‐resolution vectorial imaging of birefringent specimens, reconstructing both amplitude and quantitative phase images of two polarization channels, yielding a full Jones matrix without mechanical scanning or specialized sensors. This pupil plane multiplexing scheme offers a compact, modality‐adaptive, and scalable platform for advanced multichannel microscopy.

Electromagnetic hyperbolicity has driven key functionalities in nanophotonics, including super-resolution imaging, efficient energy control, and extreme light manipulation. Central to these advances are hyperbolic polaritons-nanometer-scale light-matter waves-spanning multiple energy-momentum dispersion orders with distinct mode profiles and incrementally high optical momenta. In this work, we report the mode conversion of hyperbolic polaritons across different dispersion orders by breaking the structure symmetry in engineered step-shape van der Waals (vdW) terraces. The mode conversion from the fundamental to high-order hyperbolic polaritons is imaged using scattering-type scanning near-field optical microscopy (s-SNOM) on both hexagonal boron nitride (hBN) and alpha-phase molybdenum trioxide (α-MoO3) vdW terraces. Our s-SNOM data, augmented with electromagnetics simulations, further demonstrate the alteration of polariton mode conversion by varying the step size of vdW terraces. The mode conversion reported here offers a practical approach toward integrating previously independent different-order hyperbolic polaritons with ultra-high momenta, paving the way for promising applications in nano-optical circuits, sensing, computation, information processing, and super-resolution imaging.

Abstract Understanding two‐phase flow in laminated sandstones is important for fluid migration control and operational strategy determination in underground energy and hydrology engineering projects. Digital core analysis provides unparalleled understanding of two‐phase flow in complex porous media, but the integration into field analytical workflow is obstructed by the huge computational burden and imaging limitations on a standard rock core. To address this challenge, we propose a novel pore‐scale rock‐typing and upscaling approach for fast computation of two‐phase flow properties on large three‐dimensional digital rock images that contain billions of voxels. Firstly, a heterogeneous rock sample is divided into several homogeneous rock types through data clustering of regional 3D morphological parameters, and their two‐phase flow properties are calculated from selected subsamples using in‐house pore‐network model. The capillary pressure and relative permeability curves of the full digital image are then estimated through quasi‐static modeling on the rock type distribution. The excellent agreement between the upscaling results and pore‐scale simulations on the full image has verified the effectiveness of this two‐phase flow upscaling strategy. With largely reduced computational demands and clearly defined lamination heterogeneity, this approach has demonstrated good potential in bridging the gap between pore‐scale and core‐scale fluid flow mechanisms. In addition, due to the laminated structural characteristics, we also find a significant reduction in phase mobility over a range of saturations in the relative permeability curves of this highly permeable rock sample.

The digitization of archival repositories has revolutionized marketing expectations, offering limitless possibilities at unparalleled speeds. Artificial Intelligence-driven content creation through neural networks has transformed the discourse on the perpetuation of ideas, concepts, and cultural heritage. Stakeholders now recognize the pivotal role of AI in leveraging archival heritage, utilizing productive tools to both market and preserve the ethnic essence of indigenous memory variables. Traditionally, the narratives of wandering aborigines were confined to oral traditions, but with the advent of computer-based technologies, these stories have found a new medium in digital archives. The intersection of computational virtual vision and generative images technology has provided patrons and researchers with unprecedented access to explore indigenous traditions and identities. This technological advancement not only preserves ethnic heritage but also serves as a gateway to sustainable ecosystems, reinvigorating intangible cultural markers within neural network datasets. Sophisticated Language Models (LLMs) play a crucial role in cross-cultural communication, facilitating the translation and interpretation of diverse cultural contexts. By analysing acoustic features, visual imagery, myths, symbols, and motifs, these models contribute significantly to multi-dimensional learning paradigms, offering insights into neurological perceptions across different cultural and linguistic landscapes. Art enthusiasts are drawn to exploring the intricate relationship between time, ecology, biology, and environmental factors. This exploration delves into past and present realities, shedding light on both intrinsic and extrinsic motivations that shape individuals' interactions with their surroundings. After studying some samples as case studies of natural language processing and neural network programming especially of the ethnography of folk culture from Australia and India, it appeared that virtual orientation is in fact and in principle a purpose of building pedagogues of virtual orientalism, besides being the resourceful neurons to calculate perceptron (a mathematical model of a biological neuron used in AI NNs or a simple algorithm to classify data) for multi-layer neural computational automated vision. Indeed, the wandering aborigines’ culture is now a wandering computational pool to build national interest for traditions and indigeneity, and to prevent their extinction, besides mitigating binaries of nature and culture. The paper aims to present an overview of the involved Repository learning models’ performance initiated to preserve and restore the process of loss, the function, and the training. Secondly the paper will also attempt to present the pro-active steps taken by the governing agencies in cross-cultural context to conserve intangible assets for generating text and content for the further academic proposed scholarships.

ABSTRACT Moving beyond structural mimicry of insect compound eyes, we present a bioinspired vision system that replicates both the optical architecture and neural processing of insect motion detection pathways. Our single‐pixel‐based artificial compound eye (SPBACE) integrates insect‐inspired optoelectronic designs with event‐based neuromorphic processing, achieving motion detection 35,714 signals per second (ss/s), surpassing conventional sensor arrays by three orders of magnitudes. The system processed multiple 1D electronic signals rather than images for direction computation, enabling robust collision detection in non‐line‐of‐sight (NLOS) scenarios without frame‐rate limitations. This dual biomimetic approach yields breakthrough performances in high‐speed motion perception, with demonstrated applications in autonomous navigation where rapid threat assessment is critical. By bridging computational imaging with bio‐inspired algorithms, we establish a paradigm for artificial vision system that combines the structural advantages of compound eyes with processing efficiency of insect nervous systems.

This paper examines the visual semantics of AI generated paintings focusing on the crossroads between the computational creativity and the aesthetic interpretation of paintings by humans. With the rise of artificial intelligence based on powerful image-generation models, including Generative Adversarial Networks (GANs) and diffusion models, the ability of machines to generate image-based artworks with visual complexity and symbolic richness has increased. Nonetheless, it is a question to whether these visual products contain actual semantic depth or they only imitate human artistic intent. The study is based on a mixed-method design that involves a combination of both computational image analysis and qualitative semantic interpretation in order to explore the construction and perception of meaning within AI-generated art. The results of analysis of a curated dataset of AI-generated paintings were presented using CLIP, DALL•E, and Midjourney to obtain visual features and project them on conceptual and emotional planes. By applying the conceptual theories of semiotics and aesthetics, the paper determines the trends in color, composition and symbolism that are used to encode cultural and perceptual information in AI models. It has been found that though AI systems are able to adopt a human-like semantics via visual correlations learned by training, their results are essentially derivational: based on training data and probability associations as opposed to capturing original creative intent. The discourse explains the ways AI-generated paintings question traditional limits of authorship and artistic meaning in that they propose a new paradigm, in which human users collaborate to create semantics out of algorithmic aesthetics.

Automated Bone Fragment Classification Using Computational Imaging and Machine Learning

The quality of retinal images is compromised by aberrations that remain uncorrected even in confocal adaptive optics imaging. This study demonstrates phase diversity (PD), a computational imaging technique, to address residual aberrations and enhance image quality in adaptive optics scanning laser ophthalmoscopy (AOSLO). By using images of the same object obtained with and without deliberately added aberrations, PD computes and compensates for the effects of existing residual aberrations beyond those corrected by a closed-loop AO system. Experimental validation demonstrates that PD improves visualization of retinal microstructures, including cone mosaics and dendrites of fluorescently labeled retinal ganglion cells (RGCs).

Blood-handling devices are commonly used for blood transportation or regulation, but their specialized flow channel geometries tend to create high-shear-stress flow regimes, which may induce excessive cellular damage risks and energy dissipation. To address this, this study combines computational fluid dynamics and particle image velocimetry experimental methods to establish nozzle reference models with multiple orifice diameter configurations. Based on entropy generation theory and Ω vortex identification methods, the underlying energy dissipation mechanisms and vortex dynamics under distinct high-shear-stress conditions are analyzed. The results indicate that shear flow intensity is highly correlated with energy dissipation due to entropy production. Attenuating turbulence in the flow field simultaneously suppresses shear stress damage and energy loss, while lowering shear flow intensity promotes the decomposition of vortices downstream, broadening their spatial distribution. High flow velocity alone does not directly induce shear stress or entropy-related energy dissipation; rather, an excessively steep velocity gradient is the primary factor affecting flow field safety and efficiency. A 94% rise in velocity gradient results in average increases of 97.6% in shear stress and 99.6% in energy entropy production. During flow regime transition or under pronounced velocity gradients, shear-dominated vortices readily form and generate vortex-like energy dissipation during evolution, which is a key factor exacerbating energy loss in high-shear-stress flow fields. This study elucidates the energy dissipation mechanisms and vortex dynamics in high-shear-stress flow fields of blood-handling devices, providing theoretical and technical support for optimizing flow fields and performance in relevant devices.

Single-photon computational ghost imaging (SCGI) typically operates in ultra-low-echo photon flux environment (ULEPFE), resulting in a low detecting signal-noise-ratio (SNR) and a complicated detecting process. Here, our aim is to develop a high-echo photon computational ghost imaging technique (HCGI) to avoid pile-up effect utilizing the pairwise orthogonality of the Hadamard matrix and improve the detection efficiency with a 64 × 64 SPAD camera. HCGI finished a 256 × 256 qualified imaging in a high-echo photon flux environment (HEPFE) at a distance of 38.31 m when operating with 200 detections per pattern, and simplifies the data processing. We also developed and validated a new imaging quality metrics formula for HCGI. Our study demonstrates that increasing pulse number, optical magnification, and attenuation coefficient considerably enhances the imaging quality. The advancements of our work in single-photon imaging are particularly beneficial for applications such as biomedical imaging, night vision, and remote sensing.

Retraction Note: Simulation of computer image recognition technology based on image feature extraction

Hybrid Liquid-Lens-Based Diffractive Spectral Array Computational Imaging with a Parallel Fused Spatially Calibrated 3D Network