High-resolution Reconstruction Research Articles

Physics-informed neural networks for high-resolution weather reconstruction from sparse weather stations.

The accurate provision of weather information holds immense significance to many disciplines. One example corresponds to the field of air traffic management, in which one basis for weather detection is set upon recordings from sparse weather stations on ground. The scarcity of data and their lack of precision poses significant challenges to achieve a detailed description of the atmosphere state at a certain moment in time. In this article, we foster the use of physics-informed neural networks (PINNs), a type of machine learning (ML) architecture which embeds mathematically accurate physics models, to generate high-quality weather information subject to the regularization provided by the Navier-Stokes equations. The application of PINNs is oriented to the reconstruction of dense and precise wind and pressure fields in areas where only a few local measurements provided by weather stations are available. Our model does not only disclose and regularize such data, which are potentially corrupted by noise, but is also able to precisely compute wind and pressure in target areas. The effect of time and spatial resolution over the capability of the PINN to accurately reconstruct fluid phenomena is thoroughly discussed through a parametric study, concluding that a proper tuning of the neural network's loss function during training is of utmost importance.

MPIGAN: An end-to-end deep based generative framework for high-resolution magnetic particle imaging reconstruction.

Magnetic particle imaging (MPI) is a recently developed, non-invasive in vivo imaging technique to map the spatial distribution of superparamagnetic iron oxide nanoparticles (SPIONs) in animal tissues with high sensitivity and speed. It is a challenge to reconstruct images directly from the received signals of MPI device due to the complex physical behavior of the nanoparticles. System matrix and X-space are two commonly used MPI reconstruction methods, where the former is extremely time-consuming and the latter usually produces blurry images. Currently, we proposed an end-to-end machine learning framework to reconstruct high-resolution MPI images from 1-D voltage signals directly and efficiently. The proposed framework, which we termed "MPIGAN", was trained on a large MPI simulation dataset containing 291597 pairs of high-resolution 2-D phantom images and each image's corresponding voltage signals, so that it was able to accurately capture the nonlinear relationship between the spatial distribution of SPIONs and the received voltage signal, and realized high-resolution MPI image reconstruction. Experiment results showed that, MPIGAN exhibited remarkable abilities in high-resolution MPI image reconstruction. MPIGAN outperformed the traditional methods of system matrix and X-space in recovering the fine-scale structure of magnetic nanoparticles' spatial distribution and achieving enhanced reconstruction performance in both visual effects and quantitative assessments. Moreover, even when the received signals were severely contaminated with noise, MPIGAN could still generate high-quality MPI images. Our study provides a promising AI solution for end-to-end, efficient, and high-resolution magnetic particle imaging reconstruction.

Noise reduction of axial piston motors using multi-objective optimization of reinforcing ribs based on identification of acoustic sources

The optimization of the housing for reducing vibration and noise in axial piston motors is facing challenges related to inaccurate locations of acoustic sources and inappropriate optimization parameters. In this study, a multi-objective optimization of the thicknesses of the reinforcing ribs for reducing the noise of axial piston motors is investigated by using the compression sensing (CS) method and response surface methodology (RSM). Combined with the CS method, the high-resolution reconstruction of sound-intensity images is used to guide the structural optimization areas through locations of acoustic sources. To reduce vibration and noise, the variable rib structure model (VRSM) is established in structural optimization areas. The thicknesses of reinforcing ribs in different areas are considered optimization variables, while the motor housing's peak vibration acceleration obtained by coupling the motor's dynamic model with the hydraulic model is utilized as the objective function. The mathematical relationship between these variables and the objective function is established through the RSM. Utilizing these functions, a multi-objective optimization model is formulated for discretizing the thicknesses of reinforcing ribs to reduce noise, employing a multi-objective genetic algorithm (MOGA) for optimization calculations. Finally, comparing the sound pressure levels of the axial piston motor with and without the optimized reinforcing ribs, the experimental results demonstrate a notable reduction in noise with the optimized reinforcing ribs. Specifically, there's a reduction of 2.7 dB at the peak frequency of 610 Hz.

New Vacuum Solar Telescope Achieves Narrowband Infrared Solar Imaging Observation at He i 10830 Å

The near-infrared imaging channel constitutes a crucial component of the multichannel high-resolution imaging system of the New Vacuum Solar Telescope (NVST). We have successfully achieved high-resolution, narrowband imaging of the chromosphere using He i 10830 Å triplet within this channel, which significantly enhances the imaging observation capabilities of NVST. This paper provides a concise overview of the optical system associated with the near-infrared imaging channel, detailing data processing procedures and presenting several observed images. Leveraging a high-resolution image reconstruction algorithm, we were able to generate a narrowband image near the diffraction limit at 10830 Å with a temporal resolution of less than 10 s.

High Resolution Reconstruction of the Proximal Tubule Apical Endocytic Pathway

Kidney proximal tubule (PT) cells are specialized for effcient apical uptake of albumin and other proteins that escape the glomerular filtration barrier. The apical endocytic pathway is uniquely organized to maintain highly dynamic retrieval of these proteins via binding to the multiligand receptors megalin and cubilin. Ultrastructural studies following the fate of internalized markers have provided a temporal model for the delivery of receptor-bound ligands and fluid phase markers to lysosomes. In this model, internalization from the apical membrane occurs via budding of irregular clathrin-coated invaginations that form at the base of microvilli. After uncoating, budded vesicles fuse with apical early endosomes (AEEs) where acidification triggers dissociation of ligands from their receptors. AEEs further mature and expand into apical vacuoles (AVs) that ultimately deliver ligands and fluid to lysosomes. Receptor recycling to the apical membrane is thought to occur via dense apical tubules (DATs) that have a uniform diameter and membrane thickness and which appear to bud from both AEEs and AVs. We used Zeiss Arivis Pro technology to annotate and reconstruct apical membrane invaginations and endocytic compartments in a 3D volume of mouse kidney with isotropic voxels of 4 nm, acquired using focused ion-beam scanning electron microscopy (FIB-SEM). Strikingly, 3D reconstruction of DATs revealed a large and highly interconnected network of tubules in the subapical region of PT cells. Connections were observed between the DAT network and endosomal compartments, as well as between DATs and apical membrane invaginations. Our data suggest the possibility that DATs exist as a stable compartment that interacts transiently with the apical membrane to enable receptor recycling. Current studies are underway to better understand how PT cells modulate endocytic entry and recycling via the same portals. National Institute of Health: 5T32GM133353-03, R01 DK118726, R01 DK125049, U54 DK137329. This is the full abstract presented at the American Physiology Summit 2024 meeting and is only available in HTML format. There are no additional versions or additional content available for this abstract. Physiology was not involved in the peer review process.

Deep learning-enhanced ghost imaging through dynamic and complex scattering media with supervised corrections of dynamic scaling factors

Ghost imaging (GI) through dynamic and complex scattering media remains challenging. The existence of dynamic scattering gives rise to a failure of GI schemes. Here, we report a deep learning-enhanced GI scheme with supervised corrections (SCGI) of dynamic scaling factors to realize high-resolution ghost reconstruction through dynamic and complex scattering media. The SCGI scheme is developed to approximate the variation of dynamic scaling factors in an optical channel and correct the recorded light intensities with a Gaussian prior. An untrained neural network powered by regularization by denoising for the SCGI scheme (SCGI-URED) is developed to further recover high-visibility ghost images. Experimental results demonstrate that high-resolution and high-visibility GI can be realized in dynamic and complex scattering media. The proposed method provides a reliable tool for implementing high-resolution and high-visibility GI through dynamic and complex scattering media and could give an impetus to developing dynamic scattering imaging in real-world scenarios.

Asian monsoon variations over the past 21 ka: An introduction

The Asian monsoon influences the lives of more than a third of the world's population, and understanding its spatial and temporal variability is fundamental to sustainable development. The complex dynamics of this climatic system are therefore of broad scientific and societal interest. In this virtual special issue (VSI), we present 19 papers dealing with the Asian monsoon since the Last Glacial Maximum (LGM, ∼21 kyr BP), with a focus on understanding variability and forcing mechanisms. Findings use recently developed quantitative reconstructions combined with numerical simulations, and present novel insights as follows: (1) analyses of comprehensive datasets of pollen, diatom, brGDGTs, oxygen and hydrogen isotopes and frequency magnetic susceptibility, as well as an innovative algorithm model for the lake energy-water balance, improve the reliability of quantitative reconstructions of Asian monsoon climate variations; (2) studies of quantitative climate parameters in central East Asia show that mean annual temperature (MAT) was ∼5 °C (or ∼ 7.9 °C using brGDGTs) lower during the LGM, ∼1.5 °C (±41% using brGDGTs) higher around the Holocene thermal optimum (HO) and mean annual precipitation (MAP) varied between 30 and 150% with marked seasonality during 8–5 ka; (3) dynamical downscaling palaeoclimatic simulations improve the accuracy of past monsoon variations, revealing strong variability in monsoon precipitation across the LGM and HO. The holistic study of Asian monsoon variations at orbital to centennial timescales in these VSI papers presents reliable observations of monsoon variability and its driving mechanisms, providing high-resolution quantitative paleoclimate reconstructions that fill existing gaps in spatial coverage, as well as offering an improved understanding, thorough numerical simulation, of coupled sea-level change and vegetation feedbacks.

Physics-informed deep-learning applications to experimental fluid mechanics

High-resolution reconstruction of flow-field data from low-resolution and noisy measurements is of interest due to the prevalence of such problems in experimental fluid mechanics, where the measurement data are in general sparse, incomplete and noisy. Deep-learning approaches have been shown suitable for such super-resolution tasks. However, a high number of high-resolution examples is needed, which may not be available for many cases. Moreover, the obtained predictions may lack in complying with the physical principles, e.g. mass and momentum conservation. Physics-informed deep learning provides frameworks for integrating data and physical laws for learning. In this study, we apply physics-informed neural networks (PINNs) for super-resolution of flow-field data both in time and space from a limited set of noisy measurements without having any high-resolution reference data. Our objective is to obtain a continuous solution of the problem, providing a physically-consistent prediction at any point in the solution domain. We demonstrate the applicability of PINNs for the super-resolution of flow-field data in time and space through three canonical cases: Burgers’ equation, two-dimensional vortex shedding behind a circular cylinder and the minimal turbulent channel flow. The robustness of the models is also investigated by adding synthetic Gaussian noise. Furthermore, we show the capabilities of PINNs to improve the resolution and reduce the noise in a real experimental dataset consisting of hot-wire-anemometry measurements. Our results show the adequate capabilities of PINNs in the context of data augmentation for experiments in fluid mechanics.

Reactive UAV-based automatic tunnel surface defect inspection with a field test

This work addresses the problem of automatic tunnel surface defect inspection using unmanned aerial vehicles (UAVs). The research aims at proposing a robust and efficient monitoring method for image data acquisition and processing in complex and dark tunnel environments. A method, called Proximity Move-Pause-Photo for Surface Defect Inspection (PMPP-SDI), is proposed by combining reactive flying control strategies with a grid scanning pattern to capture high-quality image data from multiple views and angles. The image data is then used to generate a 3D point cloud model of the tunnel surface for structural condition assessment. The method is tested in a field experiment in a railway tunnel in Ireland, and the results show that it can achieve stable navigation, high-resolution reconstruction, and accurate defect detection. The paper discusses the advantages and limitations of the method, and suggests improving the control/navigation intelligence, data quality, and defect analysis as the future research directions.

Dual-branch feature encoding framework for infrared images super-resolution reconstruction

Infrared thermal imaging is a passive non-contact detection and identification technology, which is not subject to electromagnetic infection and good concealment, is widely used in military and commercial fields. However, due to the limitations of the existing infrared imaging system mechanisms, the spatial resolution of the acquired infrared images is low and the edge details are blurred, which in turn leads to poor performance in downstream missions based on infrared images. In this paper, in order to better solve the above problems, we propose a new super-resolution reconstruction framework for infrared images, called DBFE, which extracts and retains abundant structure and textual information for robust infrared image high-resolution reconstruction with a novel structure-textual encoder module. Extensive experiment demonstrates that our proposed method achieves significantly superior contraband high-resolution reconstruction results on the multiple dataset compared to progressive methods for high resolution infrared image reconstruction, effectively proving the practicability of the method proposed in this paper.

Optical fiber bundle differential compressive imaging.

We present a differential compressive imaging method for an optical fiber bundle (OFB), which provides a solution for an ultrathin bend-resistant endoscope with high resolution. This method uses an OFB and a diffuser to generate speckle illumination patterns. Differential operation is additionally applied to the speckle patterns to produce sensing matrices, by which the correlation between the matrices is greatly reduced from 0.875 to 0.0275, which ensures the high quality of image reconstruction. Pixilation artifacts from the fiber core arrangement are also effectively eliminated with this configuration. We demonstrate high-resolution reconstruction of images of 132 × 132 pixels with a compression rate of 12% using 77 fiber cores, the total diameter of which is only about 91 µm. An experimental verification proves that this method is tolerant to a limited degree of fiber bending, which provides a potential approach for robust high-resolution fiber endoscopy.

Automated Camera Pose Generation for High-Resolution 3D Reconstruction of Bridges by Unmanned Aerial Vehicles

This work explores the possibility of automating the aerial survey of bridges to generate high-resolution images necessary for digital damage inspection. High-quality unmanned aerial vehicle (UAV) based 3D reconstruction of bridges is an important step towards autonomous infrastructure inspection. However, the calculation of optimal camera poses remains challenging due to the complex structure of bridges and is therefore often conducted manually. This process is time-consuming and can lead to quality losses. Research in this field to automate this process is yet sparse and often requires high informative models of the bridge as the base for calculations, which are not given widely. Therefore, this paper proposes an automated camera pose calculation method solely based on an easily accessible polygon mesh of the bridge. For safe operation, point cloud data of the environment are used for automated ground detection and obstacle avoidance including vegetation. First, an initial set of camera poses is generated based on a voxelized mesh created in respect to the quality requirements for 3D reconstruction using defined camera specification. Thereafter, camera poses not fulfilling safety distances are removed and specific camera poses are added to increase local coverage quality. Evaluations of three bridges show that for diverse bridge types, near-complete coverage was achieved. Due to the low computational effort of the voxel approach, the runtime was kept to a minimum, even for large bridges. The subsequent algorithm is able to find alternative camera poses even in areas where the optimal pose could not be placed due to obstacles.

Simplified matching pursuits applied to 3D nuclear reactor temperature distribution construction

Autonomous control systems can enhance the economic viability of complex systems such as nuclear microreactors. Advanced forms of these control systems benefit from techniques for high-resolution field data reconstruction from sensor data. Basis projection methods present a useful set of methodologies for nonparametric reconstruction of these field data. In this work, high fidelity multiphysics simulations are used to generate microreactor temperature distributions that are then used to evaluate the viability of a basis projection method which employs a generalized set of basis functions, or dictionaries. Additionally, metrics for evaluating the accuracy of the temperature distribution are presented that are motivated both from the fields of traditional signal processing and statistics. The performance of a method based on matching pursuits is compared to a method that uses the sequential ordering of basis functions. The matching pursuits-based method gives improved performance for highly sparse representations of temperature distributions. Additionally, accuracy is analyzed for multiple temperature distributions which represent different potential reactor operation configurations. Overall, the results from this work should be considered as an upper bound on the performance of basis projection methods which employ generalized basis functions for real time reactor control.

Reconstruction of warm-season temperatures in central Europe during the past 60 000 years from lacustrine branched glycerol dialkyl glycerol tetraethers (brGDGTs)

Abstract. Millennial-scale climate variations during the last glacial period, such as Dansgaard–Oeschger (DO) cycles and Heinrich events, have been extensively studied using ice core and marine proxy records. However, there is a limited understanding of the magnitude of these temperature fluctuations in continental regions, and questions remain about the seasonal signal of these climate events. This study presents a 60 000-year-long temperature reconstruction based on branched glycerol dialkyl glycerol tetraethers (brGDGTs) extracted from lake sediments from the Eifel Volcanic Field, Germany. brGDGTs are bacterial membrane-spanning lipids that are known to have a strong relationship with temperature, making them suitable for temperature reconstructions. We test several temperature calibration models on modern samples taken from soils and multiple maar lakes. We find a negative bias in brGDGT-based temperature estimates associated with water depth and anoxic conditions that can be corrected for by accounting for a brGDGT isomer that is only produced in anoxic conditions. The corrected temperature reconstruction correlates with proxy and climate model estimates of temperature spanning the same time period, validating the calibration approach we selected. However, millennial-scale variability is significantly dampened in the brGDGT record, and in contrast to other Northern Hemisphere climate records, during several Heinrich stadials, temperatures actually increase. We demonstrate that these apparent discrepancies can be explained by the unique seasonal response of the brGDGT paleothermometer to temperatures of months above freezing (TMAF). Our data support the view that warm-season temperatures in Europe varied minimally during the last glacial period and that abrupt millennial-scale events were defined by colder, longer winters. Our continuous high-resolution temperature reconstruction provides important information about the magnitude of seasonal climate variability during the last glacial period that can be used to test climate models and inform studies of paleoecological change.

Late Holocene high-resolution paleoenvironmental reconstruction from Beaver Lake in the northwest lowlands of the Olympic Peninsula, Washington (USA)

Fire is an essential component of the landscapes and forests of the Pacific Northwest, including the temperate rainforests of the Olympic Peninsula. Previous fire history reconstructions from the peninsula show that fire return intervals varied throughout the postglacial period, primarily in response to climatic changes and corresponding shifts in vegetation. However, much less is known about the fire history of the low-elevation forests of the Olympic Peninsula and the role of cultural fire regimes in these environments. The purpose of this study was to reconstruct the paleoenvironmental history of a low-elevation study site, Beaver Lake, located in the northwestern part of the peninsula. Using macroscopic charcoal, pollen, and sedimentological analyses of a ca. 3440-year-long record, we reconstructed a high-resolution record that shows fire activity was remarkably high during the late Holocene for an area considered part of a temperate rainforest. However, patterns of burning varied throughout the record, with the first third (ca. 3440 to 2350 cal yr BP) and last third (ca. 800 cal yr BP to the present day) recording much higher amounts of fire activity compared to the middle portion (ca. 2350 to 800 cal yr BP), which recorded very little fire activity. Larger and/or more severe fires that likely burned during multi-year droughts correlate with peaks in magnetic susceptibility and pollen trends that indicate substantial geomorphic responses and successional changes in forest structure following these events. Cooler temperatures and a wetter climate, indicated by nearby glacial advancements, may have staved off fire activity during the period of low fire activity. Cultural burning by the Indigenous tribes of the Olympic Peninsula likely contributed to the observed fire activity at Beaver Lake as nearby prairies were managed for resource procurement, indicated in particular by an increase in herbaceous morphotype charcoal during the past 800 years.

Impact of PET Reconstruction on Amyloid-β Quantitation in Cross-Sectional and Longitudinal Analyses.

Amyloid-β (Aβ) accumulation in Alzheimer disease (AD) is typically measured using SUV ratio and the centiloid (CL) scale. The low spatial resolution of PET images is known to degrade quantitative metrics because of the partial-volume effect. This article examines the impact of spatial resolution, as determined by the reconstruction configuration, on the Aβ PET quantitation in both cross-sectional and longitudinal data. Methods: The cross-sectional study involved 89 subjects with 20-min [18F]florbetapir scans generated on an mCT (44 Aβ-negative [Aβ-], 45 Aβ-positive [Aβ+]) using 69 reconstruction configurations, which varied in number of iteration updates, point-spread function, time-of-flight, and postreconstruction smoothing. The subjects were classified as Aβ- or Aβ+ visually. For each reconstruction, Aβ CL was calculated using CapAIBL, and the spatial resolution was calculated as full width at half maximum (FWHM) using the barrel phantom method. The change in CLs and the effect size of the difference in CLs between Aβ- and Aβ+ groups with FWHM were examined. The longitudinal study involved 79 subjects (46 Aβ-, 33 Aβ+) with three 20-min [18F]flutemetamol scans generated on an mCT. The subjects were classified as Aβ- or Aβ+ using a cutoff CL of 20. All scans were reconstructed using low-, medium-, and high-resolution configurations, and Aβ CLs were calculated using CapAIBL. Since linear Aβ accumulation was assumed over a 10-y interval, for each reconstruction configuration, Aβ accumulation rate differences (ARDs) between the second and first periods were calculated for all subjects. Zero ARD was used as a consistency metric. The number of Aβ accumulators was also used to compare the sensitivity of CL across reconstruction configurations. Results: In the cross-sectional study, CLs in both the Aβ- and the Aβ+ groups were impacted by the FWHM of the reconstruction method. Without postreconstruction smoothing, Aβ- CLs increased for a FWHM of 4.5 mm or more, whereas Aβ+ CLs decreased across the FWHM range. High-resolution reconstructions provided the best statistical separation between groups. In the longitudinal study, the median ARD of low-resolution reconstructed data for the Aβ- group was greater than zero whereas the ARDs of higher-resolution reconstructions were not significantly different from zero, indicating more consistent rate estimates in the higher-resolution reconstructions. Higher-resolution reconstructions identified 10 additional Aβ accumulators in the Aβ- group, resulting in a 22% increased group size compared with the low-resolution reconstructions. Higher-resolution reconstructions reduced the average CLs of the negative group by 12 points. Conclusion: High-resolution PET reconstructions, inherently less impacted by partial-volume effect, may improve Aβ PET quantitation in both cross-sectional and longitudinal data. In the cross-sectional analysis, separation of CLs between Aβ- and Aβ+ cohorts increased with spatial resolution. Higher-resolution reconstructions also exhibited both improved consistency and improved sensitivity in measures of Aβ accumulation. These features suggest that higher-resolution reconstructions may be advantageous in early-stage AD therapies.

AcquisitionFocus: Joint Optimization of Acquisition Orientation and Cardiac Volume Reconstruction Using Deep Learning.

In cardiac cine imaging, acquiring high-quality data is challenging and time-consuming due to the artifacts generated by the heart's continuous movement. Volumetric, fully isotropic data acquisition with high temporal resolution is, to date, intractable due to MR physics constraints. To assess whole-heart movement under minimal acquisition time, we propose a deep learning model that reconstructs the volumetric shape of multiple cardiac chambers from a limited number of input slices while simultaneously optimizing the slice acquisition orientation for this task. We mimic the current clinical protocols for cardiac imaging and compare the shape reconstruction quality of standard clinical views and optimized views. In our experiments, we show that the jointly trained model achieves accurate high-resolution multi-chamber shape reconstruction with errors of <13 mm HD95 and Dice scores of >80%, indicating its effectiveness in both simulated cardiac cine MRI and clinical cardiac MRI with a wide range of pathological shape variations.

Clumped isotopes reveal relationship between mussel growth and river discharge

Freshwater mussels preserve valuable information about hydrology, climate, and population dynamics, but developing seasonal chronologies can be problematic. Using clumped isotope thermometry, we produced high-resolution reconstructions of modern and historic (~ 1900) temperatures and δ18Owater from mussel shells collected from an impounded river, the Brazos in Texas, before and after damming. We also performed high-resolution growth band analyses to investigate relationships between mussel growth rate, rainfall, and seasonal temperature. Reconstructed δ18Owater and temperature vary little between the modern (3R5) and historic shell (H3R). However, a positive relationship between reconstructed δ18Owater and growth rate in H3R indicates that aside from diminished growth in winter, precipitation and flow rate are the strongest controls on mussel growth in both modern and pre-dam times. Overall, our results demonstrate (1) the impact, both positive and negative, of environmental factors such as flow alteration and temperature on mussel growth and (2) the potential for clumped isotopes in freshwater mussels as a paleohydrology and paleoclimate proxies in terrestrial environments.

Quantitative susceptibility mapping through model-based deep image prior (MoDIP)

The data-driven approach of supervised learning methods has limited applicability in solving dipole inversion in Quantitative Susceptibility Mapping (QSM) with varying scan parameters across different objects. To address this generalization issue in supervised QSM methods, we propose a novel training-free model-based unsupervised method called MoDIP (Model-based Deep Image Prior). MoDIP comprises a small, untrained network and a Data Fidelity Optimization (DFO) module. The network converges to an interim state, acting as an implicit prior for image regularization, while the optimization process enforces the physical model of QSM dipole inversion. Experimental results demonstrate MoDIP's excellent generalizability in solving QSM dipole inversion across different scan parameters. It exhibits robustness against pathological brain QSM, achieving over 32 % accuracy improvement than supervised deep learning methods. It is also 33 % more computationally efficient and runs 4 times faster than conventional DIP-based approaches, enabling 3D high-resolution image reconstruction in under 4.5 min.

LMD: Faster Image Reconstruction with Latent Masking Diffusion

As a class of fruitful approaches, diffusion probabilistic models (DPMs) have shown excellent advantages in high-resolution image reconstruction. On the other hand, masked autoencoders (MAEs), as popular self-supervised vision learners, have demonstrated simpler and more effective image reconstruction and transfer capabilities on downstream tasks. However, they all require extremely high training costs, either due to inherent high temporal-dependence (i.e., excessively long diffusion steps) or due to artificially low spatial-dependence (i.e., human-formulated high mask ratio, such as 0.75). To the end, this paper presents LMD, a faster image reconstruction framework with Latent Masking Diffusion. First, we propose to project and reconstruct images in latent space through a pre-trained variational autoencoder, which is theoretically more efficient than in the pixel-based space. Then, we combine the advantages of MAEs and DPMs to design a progressive masking diffusion model, which gradually increases the masking proportion by three different schedulers and reconstructs the latent features from simple to difficult, without sequentially performing denoising diffusion as in DPMs or using fixed high masking ratio as in MAEs, so as to alleviate the high training time-consumption predicament. Our approach allows for learning high-capacity models and accelerate their training (by 3x or more) and barely reduces the original accuracy. Inference speed in downstream tasks also significantly outperforms the previous approaches.