Fine Details Research Articles

FLUORESCENCE DETECTION OF DNA/RNA G-QUADRUPLEXES (G4s) BY TWICE-AS-SMART LIGANDS.

Fluorescence detection of DNA and RNA G-quadruplexes (G4s) is a very efficient strategy to assess not only the existence and prevalence of cellular G4s but also their relevance as targets for therapeutic interventions. Among the fluorophores used to this end, turn-on probes are the most interesting since their fluorescence is triggered only upon interaction with their G4 targets, which ensures a high sensitivity and selectivity of detection. We reported on a series of twice-as-smart G4 probes, which are both smart G4 ligands (whose structure is reorganized upon interaction with G4s) and smart fluorescent probes (whose fluorescence is turned on upon interaction with G4s). The fine mechanistic details behind the excellent properties of the best prototype N-TASQ remain to be deciphered. To investigate this, we report here on the synthesis and studies of two analogues, TzN-TASQ and AlkN-TASQ: a careful analysis of their G4-interacting properties, investigated both in vitro and in silico, shows that fine-tuning their constitutive structural elements allows for increasing the efficiency of both their 'off' (i.e., a conformation with a low fluorescence) and 'on' states (i.e., a conformation with a high fluorescence), thereby opening interesting ways for the design of more efficient fluorogenic G4 probes.

Fortunate molecules boost signal to background ratio and localization precision in correlation based single molecule localization microscopy

AbstractSingle-molecule localization microscopy (SMLM) can decipher fine details that are otherwise impossible using diffraction-limited microscopy. Often, the reconstructed super-resolved images suffer from noise, strong background and are prone to false detections that may impact quantitative imaging. To overcome these limitations, we propose a technique (corrSMLM) that recognizes and detects fortunate molecules (molecules with long blinking cycles) from the recorded data. The method uses correlation between two or more consecutive frames to identify and isolate fortunate molecules that blink longer than the standard blinking period of a molecule. The corrSMLM is based on the fact that random fluctuations (noise) do not last longer (usually limited to a single frame). In contrast, fortunate molecules consistently fluoresce for extended periods and hence appear on more than one frame. Accordingly, strongly correlated spots (representing fortunate molecules) are compared in the consecutive frames, followed by data integration to determine their position and localization precision. The technique addresses two significant problems that plague existing SMLM : (1) false detection due to random noise that contributes to a strong background and (2) poor localization leading to overall low resolution. To demonstrate, corrSMLM is used for imaging fixed NIH3T3 cells (transfected with Dendra2-Actin, Dendra2-Tubulin, and mEos-Tom20 plasmid DNA). The super-resolved images show a significant reduction in background noise ( > 1.5 fold boost in SBR) and > 2-fold improvement in localization precision as compared to standard SMLM. Intensity analysis based on the number of molecules suggests that corrSMLM better corroborates the raw data and preserves finer features (e.g., edges), which are wiped out in standard SMLM. Overall, an improvement is noted in the localization precision and spatial resolution. The proposed technique is anticipated to advance SMLM and is expected to contribute to a better understanding of single-molecule dynamics in a cellular environment.

Spectral resolution effects on the information content in solar spectra

When interpreting spectropolarimetric observations of the solar atmosphere, wavelength variations in the emergent intensity and polarization translate into information on the depth stratification of physical parameters such as the temperature, velocity, and magnetic field. Resolving the fine details in the shapes of the spectral lines and their polarization gives us the capability to resolve small-scale depth variations in these physical parameters. With the advent of large-aperture solar telescopes and the development of state-of-the-art instrumentation, the requirements on spectral resolution have become a prominent question. We aim to quantify how the information content contained in a representative set of polarized spectra of photospheric spectral lines depends on the spectral resolution and spectral sampling of that spectrum. We used a state-of-the-art numerical simulation of a sunspot and the neighboring quiet Sun photosphere to synthesize polarized spectra of magnetically sensitive neutral iron lines. We then applied various degrees of spectral degradation to the synthetic spectra and analyzed the impact on its dimensionality using principal component analysis and the wavelength power spectrum using wavelet decomposition. Finally, we applied the Stokes Inversion based on Response functions (SIR) code to the degraded synthetic data to assess the effect of spectral resolution on the inferred parameters. We find that the dimensionality of the Stokes spectra and the power contained in the small spectral scales significantly change with the spectral resolution. We find that regions with strong magnetic fields where convection is suppressed have more homogeneous atmospheres and produce less complex Stokes profiles. On the other hand, regions with strong gradients in the physical quantities give rise to more complex Stokes profiles that are more affected by spectral degradation. The degradation also makes the inversion problem more ill-defined, so inversion models with a larger number of free parameters overfit and give wrong estimates. The impact of spectral degradation in the interpretation of solar spectropolarimetric observations depends on multiple factors, including the spectral resolution, noise level, line spread function (LSF) shape, complexity of the solar atmosphere, and degrees of freedom in our inversion methods. To mitigate this impact, incorporating a good estimation of the LSF into the inversion process is recommended. Having a finely sampled spectrum may be more beneficial than achieving a higher signal-to-noise ratio per wavelength bin. Considering the inclusion of different spectral lines that can counter these effects, and calibrating the effective degrees of freedom in modeling strategies, are also important considerations. These strategies are crucial for the accurate interpretation of such observations and have the potential to offer more cost-effective solutions.

Application of the reverberation waveform inversion for the high-resolution sediment layer structure in the Pacific

Summary This study presents the velocity structures of the seafloor sediment across the Pacific using teleseismic body waves recorded by arrays of ocean bottom seismometers (OBSs). We apply the method inverting the multi-component stacked autocorrelation functions and radial-component waveform of teleseismic body waves. We analyze OBSs distributed across the Pacific Ocean, through a wide range of water depths and equipped with various sensor types. The inverted models agreeing with measurements of seafloor drilling samples demonstrate the broad applicability of employing the inversion method to resolve sediment thicknesses ranging from <100 m to a few kilometers. In addition to various broadband OBSs, we confirm our method works also for short-period sensors, since our retrieved the model using data from a short-period sensor in Northeast Japan agrees with active-source seismic survey analyses. The constrained models show various numbers of intra-sediment discontinuities across different locations. Furthermore, the newly characterized sediment models have higher values of P-wave to S-wave velocity ratio (${V}_P$/${V}_S$) than the previously known scaling relationships, which suggests a larger effect of sediment-layer reverberation on the body-wave analysis than we have anticipated. The sediment velocity models obtained in this study produce Rayleigh-wave admittances that are consistent with those in previous studies but with much finer structural details within the sediment layer. The reverberation waveform inversion method will enable improved predictions of the impact of the water and sediment layers in body-wave records from existing and future OBS observations.

Production Cross-Section of γ-Rays from (p,p′γ) Reactions: Measurements and Theoretical Analysis

This paper reviews our work on the measurements of absolute production cross sections of γ-rays from the (p,p′γ) reactions on 12C and 16O. The measurements cover a range of 8–16 MeV for the incident proton beam. The angular distributions of the γ-rays have been measured. A detailed phenomenological analysis within the framework of optical model formalism has been carried out to reproduce the experimental data. The existing global set of elastic, polarization and total reaction data for protons and neutrons have been used to generate the optical model potential. The nuclear structure effects have been included in the calculations by considering the roles of coupling of the low-lying states, the presence of resonances and nuclear deformations. The potentials so generated have been used to calculate the differential and total cross sections for both (p,p′) and (p,p′γ) reactions. The results of the analysis are in good agreement with the measured data for the observed γ-rays. However, discrepancies still exist in reproducing the finer details of the cross sections. The existing discrepancies between our phenomenological analysis and the experimental data demonstrate the rather complex roles of channel couplings, resonances in the compound nuclear system and target deformation. The significant contribution of nuclear structure effects in light mass nucleus like 12C and 16O, leads to an apparent loss of predictive power of the theoretical calculation for low-energy region (less than 10 MeV) of the projectile energy.

Features of Generation, Propagation and Application of Special Ultrasonic Impulses in Viscous Liquids

An exact numerical and approximate analytical description of solitary acoustic pulses with a large difference in spatial gradients of parameters in different directions has been obtained in viscous liquids using this small parameter. The method of special initial-boundary conditions obtained during analyzing the hydrodynamic equations has been applied to describe the peculiarities of this nonlinear phenomenon. Waves of this type exist in the presence of two- or three-dimensional inhomogeneity of the initial disturbances and retain a spatial structure along the direction of propagation when traveling long distances. At the same time, it is possible to regulate the pressure drop and the speed of the acoustic signal, which creates unique conditions for a special force effect or information transmission. The efficiency of their use in such processes as metal dissolution, solvent extraction and mass transfer under the conditions of resonance exposure of ultrasound was evaluated. Fine details of exciting the nonlinear impulse with the necessary properties have been analyzed to demonstrate a possible way to a new technology of successfully treating any different specimens, materials and constructions for a long distance between the source of radiation and the position of the treatment. The use of such pulses opens up new opportunities for remote acoustic force impact on various objects, as well as for the transmission of information.

Analysing Historical Paintings with Indian Regional Influences

Abstract: Indian paintings have a long and illustrious history that is firmly rooted in the history of Indian art. Even though there aren't many surviving early examples due to the difficult climate, the importance of paintings has persisted throughout the ancient, mediaeval, and modern periods. Indian paintings' origins can be traced back to the prehistoric era, starting with the age-old art of painting in caves. Ceramics, textiles, miniature paintings, and other media have all contributed to this rich legacy's evolution before ultimately finding expression in contemporary artistic forms. Indian painting's history, from the mysterious cave paintings to the fine details of miniature paintings, captures the spirit of the country's creative and cultural development and represents a significant trip through time. As evidenced by the remains found in caves and other textual sources, painting has been a popular art form in India since prehistoric times. The two main categories of Indian paintings are miniatures and murals. Books, paper, and linen are used for miniature paintings, whereas murals are massive pieces painted on walls and ceilings. The evolution of Indian art is a continuous process, resulting from the blending of various traditions. Indian painting has managed to hold onto its unique style while embracing contemporary trends.

Study of Penetration Depth According to Working Parameters in Steel Engraving Using Optical Laser

Laser engraving is a process in which a laser beam is used to mark or cut a material. It is a non-contact process in which the laser energy is focused and directed at the surface of the material, producing a chemical or physical change in the material. The process is fast and accurate and can be used on a wide variety of materials, including wood, metal, glass, plastic, leather, etc. The process has a variety of applications. In the field of moulds, it is used to mark moulds, to engrave company logos and certain informational text, or to engrave certain fine details on the active surface of the mould. In this paper, the use of factorial experiments is proposed to identify the factors that influence the engraving depth.

Rapid submillimeter QSM and R2* mapping using interleaved multishot 3D-EPI at 7 and 3 Tesla.

To explore the high signal-to-noise ratio (SNR) efficiency of interleaved multishot 3D-EPI with standard image reconstruction for fast and robust high-resolution whole-brain quantitative susceptibility (QSM) and mapping at 7 and 3T. Single- and multi-TE segmented 3D-EPI is combined with conventional CAIPIRINHA undersampling for up to 72-fold effective gradient echo (GRE) imaging acceleration. Across multiple averages, scan parameters are varied (e.g., dual-polarity frequency-encoding) to additionally correct for -induced artifacts, geometric distortions and motion retrospectively. A comparison to established GRE protocols is made. Resolutions range from 1.4 mm isotropic (1 multi-TE average in 36 s) up to 0.4 mm isotropic (2 single-TE averages in approximately 6 min) with whole-head coverage. Only 1-4 averages are needed for sufficient SNR with 3D-EPI, depending on resolution and field strength. Fast scanning and small voxels together with retrospective corrections result in substantially reduced image artifacts, which improves susceptibility and mapping. Additionally, much finer details are obtained in susceptibility-weighted image projections through significantly reduced partial voluming. Using interleaved multishot 3D-EPI, single-TE and multi-TE data can readily be acquired 10 times faster than with conventional, accelerated GRE imaging. Even 0.4 mm isotropic whole-head QSM within 6 min becomes feasible at 7T. At 3T, motion-robust 0.8 mm isotropic whole-brain QSM and mapping with no apparent distortion in less than 7 min becomes clinically feasible. Stronger gradient systems may allow for even higher effective acceleration rates through larger EPI factors while maintaining optimal contrast.

Credit risk assessment method driven by asymmetric loss function

Credit risk assessment is significantly hindered by the problem of class imbalance, and cost-sensitive methods represent an effective strategy to address this issue. However, most algorithms tend to approach the imbalance from a class perspective, overlooking the finer details at the sample level. Moreover, such methods are susceptible to interference from noise and outliers. In response to these challenges, this paper proposes an asymmetric cost-sensitive support vector machine (QTSVM) that combines the quadratic type squared error loss function (QTSELF) with support vector machine (SVM). It not only leverages the asymmetry of the loss function by applying varying penalties based on classification errors but also employs different processing measures for samples from different classes. Additionally, it enhances model robustness by imposing a tiny penalty on noise or outliers. The adaptive moment estimation (Adam) algorithm is employed to optimize QTSVM. Extensive experiments and statistical tests profoundly demonstrate the superior performance of QTSVM.

The Formation Mechanism of Zein Gel Induced by Amyloid-like Fibrils

The preparation of zein gels often leads to unstable product quality due to poor reproducibility, posing substantial challenges for the food industry and biomedical applications. Thus, it is essential to study the gelation mechanism and key factors responsible for gelation. Herein, we report four types of assembled structures from zein solutions in binary ethanol–water solvents with either a high concentration (30% w/v) and a low concentration (5% w/v), namely, spheres, spheres with tangential structures, amyloid-like fibrils, and sub-agglomerates with crystal structures. The tangential structure at the edges of the spheres was first observed by transmission electron microscopy (TEM) at the initial stage, offering a model for studying protein folding and aggregation mechanisms. By observing the gelation processes with TEM for morphology, cryo-TEM to confirm structure preservation, HR-TEM for fine details, and negatively stained TEM to reveal hidden features, along with small-angle X-ray scattering (SAXS), we identified two types of gels: one was mainly network structures formed by sphere–sphere fusion, which was thermoreversible, and the other one was irreversible gel due to the increase in the amount of amyloid-like fibrils. Results show that the existence of thermoreversible zein gels indicates potential effective regulation tools of zein gel properties. In addition, amyloid-like fibrils were first isolated from native zein. During amyloid-like fibril formation, the number of tangential structures increased considerably, which were composed of thinner fibrils. High-resolution TEM showed a substructure with a 0.35 nm interplanar spacing, providing evidence of β-sheet aggregates also as confirmed by SAXS. This unique substructure has not been reported in other edible proteins. Moreover, understanding the fibrillation mechanism of zein provides a basis for controlling and developing novel food structures for food protein function and offers insights into neurodegenerative diseases.

DebSDF: Delving Into the Details and Bias of Neural Indoor Scene Reconstruction.

In recent years, the neural implicit surface has emerged as a powerful representation for multi-view surface reconstruction due to its simplicity and State-of-the-Art performance. However, reconstructing smooth and detailed surfaces in indoor scenes from multi-view images presents unique challenges. Indoor scenes typically contain large texture-less regions, making the photometric loss unreliable for optimizing the implicit surface. Previous work utilizes monocular geometry priors to improve the reconstruction in indoor scenes. However, monocular priors often contain substantial errors in thin structure regions due to domain gaps and the inherent inconsistencies when derived independently from different views. This paper presents DebSDF to address these challenges, focusing on the utilization of uncertainty in monocular priors and the bias in SDF-based volume rendering. We propose an uncertainty modeling technique that associates larger uncertainties with larger errors in the monocular priors. High-uncertainty priors are then excluded from optimization to prevent bias. This uncertainty measure also informs an importance-guided ray sampling and adaptive smoothness regularization, enhancing the learning of fine structures. We further introduce a bias-aware signed distance function to density transformation that takes into account the curvature and the angle between the view direction and the SDF normals to reconstruct fine details better. Our approach has been validated through extensive experiments on several challenging datasets, demonstrating improved qualitative and quantitative results in reconstructing thin structures in indoor scenes, thereby outperforming previous work.

Impact of Deprivation and Preferential Usage on Functional Connectivity Between Early Visual Cortex and Category-Selective Visual Regions.

Human behavior can be remarkably shaped by experience, such as the removal of sensory input. Many studies of conditions such as stroke, limb amputation, and vision loss have examined how removal of input changes brain function. However, an important question yet to be answered is: when input is lost, does the brain change its connectivity to preferentially use some remaining inputs over others? In individuals with healthy vision, the central portion of the retina is preferentially used for everyday visual tasks, due to its ability to discriminate fine details. When central vision is lost in conditions like macular degeneration, peripheral vision must be relied upon for those everyday tasks, with some portions receiving "preferential" usage over others. Using resting-state fMRI collected during total darkness, we examined how deprivation and preferential usage influence the intrinsic functional connectivity of sensory cortex by studying individuals with selective vision loss due to late stages of macular degeneration. Specifically, we examined functional connectivity between category-selective visual areas and the cortical representation of three areas of the retina: the lesioned area, a preferentially used region of the intact retina, and a non-preferentially used region. We found that cortical regions representing spared portions of the peripheral retina, regardless of whether they are preferentially used, exhibit plasticity of intrinsic functional connectivity in macular degeneration. Cortical representations of spared peripheral retinal locations showed stronger connectivity to MT, a region involved in processing motion. These results suggest that the long-term loss of central vision can produce widespread effects throughout spared representations in early visual cortex, regardless of whether those representations are preferentially used. These findings support the idea that connections to visual cortex maintain the capacity for change well after critical periods of visual development.

UNet-Att: a self-supervised denoising and recovery model for two-photon microscopic image

Two-photon microscopy is indispensable in cell and molecular biology for its high-resolution visualization of cellular and molecular dynamics. However, the inevitable low signal-to-noise conditions significantly degrade image quality, obscuring essential details and complicating morphological analysis. While existing denoising methods such as CNNs, Noise2Noise, and DeepCAD serve broad applications in imaging, they still have limitations in preserving texture structures and fine details in two-photon microscopic images affected by complex noise, particularly in sophisticated structures like neuronal synapses. To improve two-photon microscopy image denoising effectiveness, by experimenting on real two-photon microscopy images, we propose a novel deep learning framework, the UNet-Att model, which integrates a specifically tailored UNet++ architecture with attention mechanisms. Specifically, this approach consists of a sophisticated downsampling module for extracting hierarchical features at varied scales, and an innovative attention module that strategically emphasizes salient features during the integration process. The architecture is completed by an ingenious upsampling pathway that reconstructs the image with high fidelity, ensuring the retention of textural integrity. Additionally, the model supports end-to-end training, optimizing its denoising efficacy. The UNet-Att model proves to surpass mainstream algorithms in the dual objectives of denoising and preserving the textural intricacies of original images, which is evidenced by an increase of 9.42 dB in the high Peak Signal-to-Noise Ratio (PSNR) coupled with an improvement of 0.1131 in the Structural Similarity Index Measurement (SSIM). The ablation experiments reveal the effectiveness of each module. The associated Python packages and datasets of UNet-Att are freely available at https://github.com/ZjjDh/UNet-Att.

GAN-Based Map Generation Technique of Aerial Image Using Residual Blocks and Canny Edge Detector

As the significance of meticulous and precise map creation grows in modern Geographic Information Systems (GISs), urban planning, disaster response, and other domains, the necessity for sophisticated map generation technology has become increasingly evident. In response to this demand, this paper puts forward a technique based on Generative Adversarial Networks (GANs) for converting aerial imagery into high-quality maps. The proposed method, comprising a generator and a discriminator, introduces novel strategies to overcome existing challenges; namely, the use of a Canny edge detector and Residual Blocks. The proposed loss function enhances the generator’s performance by assigning greater weight to edge regions using the Canny edge map and eliminating superfluous information. This approach enhances the visual quality of the generated maps and ensures the accurate capture of fine details. The experimental results demonstrate that this method generates maps of superior visual quality, achieving outstanding performance compared to existing methodologies. The results show that the proposed technology has significant potential for practical applications in a range of real-world scenarios.

Optimizing Battery Development: Integrating Air-Controlled and Cryogenic Electron Microscopy Workflows

The global energy landscape is undergoing a profound transformation, shifting from traditional fossil fuel-based systems to sustainable, renewable energy sources like wind, solar, bioenergy, hydroelectric, and tidal energy. This transition signifies a significant evolution in global energy generation and consumption patterns, necessitating advanced energy storage solutions to manage the intermittent nature of renewables and ensure a stable and reliable energy supply. Advanced energy storage technologies, including batteries, supercapacitors, and other innovative systems, play a pivotal role in this transition by enabling effective load balancing, supporting the electrification of transportation through electric vehicles and charging infrastructure, managing peak demand, and facilitating the seamless integration of renewable energy into the grid. This evolution not only advances sustainability goals but also enhances energy security and independence. [1]Advanced analytical techniques are essential for developing energy storage systems, especially batteries, by providing critical insights into their complex chemical and physical processes. Battery materials' microstructure, composition, and operational dynamics can be explored using analytical techniques like scanning electron microscopy (SEM), focused ion beam (FIB), transmission electron microscopy (TEM), X-ray diffraction (XRD), and spectroscopy. The detailed analysis through these techniques is crucial for optimizing material properties, enhancing energy density, improving safety, and extending battery lifespan. By identifying degradation mechanisms and optimizing electrode and electrolyte materials, these techniques contribute to precise battery engineering. As demand grows for more efficient and higher-capacity batteries, these sophisticated tools are indispensable for advancing energy storage technology. [2]At Shell Technology Center Houston, we have developed air-controlled and cryogenic electron microscopy (EM) workflows for studying advanced energy storage systems in preparation for the energy transition. The air-controlled EM workflow, employing SEM, FIB, TEM, and XRD, plays a pivotal role in battery research by providing a controlled environment that prevents exposure to atmospheric elements that could alter material properties. These techniques are essential for examining the fine structural details and chemical compositions of battery components under near-native conditions. The cryo-EM workflow using SEM, FIB, and TEM is vital in battery research for studying materials and interfaces sensitive to environmental conditions or rapid degradation upon exposure. These cryogenic techniques involve rapid freezing of battery components to preserve their structure in a close-to-native state and prevent deleterious effects of air exposure such as oxidation or moisture interaction. [3]In our study, we present comprehensive results obtained using both air-controlled EM and cryogenic EM workflows to investigate advanced energy storage systems, including lithium-ion batteries (LIBs) and beyond-LIBs such as multivalent rechargeable batteries. The utilization of air-controlled EM techniques such as SEM, FIB, TEM, and XRD allows us to examine battery components under controlled environmental conditions, ensuring accurate analysis of structural details and chemical compositions without the risk of atmospheric contamination. Additionally, the use of cryo-EM techniques, including cryo-SEM, cryo-FIB, and cryo-TEM, enables the study of sensitive battery materials and interfaces in a preserved, near-native state. These cryogenic methods prevent degradation effects such as oxidation and moisture interaction, providing unprecedented insights into the ultrastructural properties and operational dynamics of battery materials at atomic or molecular levels. By leveraging both air-controlled and cryogenic EM workflows, we have achieved a comprehensive understanding of battery mechanics and chemistry, paving the way for improved battery design and enhanced performance in energy storage applications.Reference[1] International Renewable Energy Agent (IRENA) 2021 https://www.irena.org/publications/2021/Aug/Renewable-energy-statistics-2021[2] Tarascon, J.-M., Armand, M., Nature 414, 359–367 (2001). https://doi.org/10.1038/35104644[3] Cheng, D., Lu, B., Raghavendran, G., Zhang, M., and Meng, Y. S., Matter, Volume 5, Issue 1, p26-42, 2022 https://doi.org/10.1016/j.matt.2021.11.019

Diffusion kurtosis imaging, MAP-MRI and NODDI selectively track gray matter myelin density in the primate cerebral cortex

Abstract Diffusion magnetic resonance imaging (dMRI) has been widely used to model the trajectory of myelinated fiber bundles in the white matter. Increasingly, it is also used to evaluate the microstructure of the cerebral cortex gray matter. For example, in diffusion tensor imaging (DTI) of the cortex, fractional anisotropy (FA) correlates strongly with the anisotropy of cellular anatomy, while radial diffusivity (RD) tracks the anisotropy of myelinated fibers. However, no DTI parameter shows specificity to gray matter myelin density. Here, we show that three higher-order diffusion parameters—the mean diffusion kurtosis (MK), the Neurite Density Index (NDI) from neurite orientation dispersion and density imaging (NODDI), and the Non-Gaussian (NG) parameter from mean apparent propagator (MAP)-MRI—each track the laminar and regional myelin density of the primate cerebral cortex in fine detail. We carried out ultra-high-resolution, multi-shelled dMRI in ex-vivo marmoset monkey brains. We compared the spatial mapping of the MK, NDI, and ND diffusion parameters to the cortical myelin distribution of these brains, with the latter obtained in two ways: First, using histological sections finely co-registered to the MRI, and second using magnetization transfer ratio MRI scans (MTR), an established non-diffusion method for imaging myelin density. We found that, in contrast to DTI parameters, each of these higher-order diffusion measures captured the spatial variation of myelin density in the cortex. The demonstration that diffusion parameters exhibit both sensitivity and specificity for gray matter myelin density will allow dMRI to more effectively track human disease, in which myelinated and non-myelinated tissue compartments are affected differentially.

MV2MV: Multi-View Image Translation via View-Consistent Diffusion Models

Image translation has various applications in computer graphics and computer vision, aiming to transfer images from one domain to another. Thanks to the excellent generation capability of diffusion models, recent single-view image translation methods achieve realistic results. However, directly applying diffusion models for multi-view image translation remains challenging for two major obstacles: the need for paired training data and the limited view consistency. To overcome the obstacles, we present a first unified multi-view image to multi-view image translation framework based on diffusion models, called MV2MV. Firstly, we propose a novel self-supervised training strategy that exploits the success of off-the-shelf single-view image translators and the 3D Gaussian Splatting (3DGS) technique to generate pseudo ground truths as supervisory signals, leading to enhanced consistency and fine details. Additionally, we propose a latent multi-view consistency block, which utilizes the latent-3DGS as the underlying 3D representation to facilitate information exchange across multi-view images and inject 3D prior into the diffusion model to enforce consistency. Finally, our approach simultaneously optimizes the diffusion model and 3DGS to achieve a better trade-off between consistency and realism. Extensive experiments across various translation tasks demonstrate that MV2MV outperforms task-specific specialists in both quantitative and qualitative.

ToonCrafter: Generative Cartoon Interpolation

We introduce ToonCrafter, a novel approach that transcends traditional correspondence-based cartoon video interpolation, paving the way for generative interpolation. Traditional methods, that implicitly assume linear motion and the absence of complicated phenomena like dis-occlusion, often struggle with the exaggerated non-linear and large motions with occlusion commonly found in cartoons, resulting in implausible or even failed interpolation results. To overcome these limitations, we explore the potential of adapting live-action video priors to better suit cartoon interpolation within a generative framework. ToonCrafter effectively addresses the challenges faced when applying live-action video motion priors to generative cartoon interpolation. First, we design a toon rectification learning strategy that seamlessly adapts live-action video priors to the cartoon domain, resolving the domain gap and content leakage issues. Next, we introduce a dual-reference-based 3D decoder to compensate for lost details due to the highly compressed latent prior spaces, ensuring the preservation of fine details in interpolation results. Finally, we design a flexible sketch encoder that empowers users with interactive control over the interpolation results. Experimental results demonstrate that our proposed method not only produces visually convincing and more natural dynamics, but also effectively handles dis-occlusion. The comparative evaluation demonstrates the notable superiority of our approach over existing competitors. Code and model weights are available at https://doubiiu.github.io/projects/ToonCrafter

Generative Portrait Shadow Removal

We introduce a high-fidelity portrait shadow removal model that can effectively enhance the image of a portrait by predicting its appearance under disturbing shadows and highlights. Portrait shadow removal is a highly ill-posed problem where multiple plausible solutions can be found based on a single image. For example, disentangling complex environmental lighting from original skin color is a non-trivial problem. While existing works have solved this problem by predicting the appearance residuals that can propagate local shadow distribution, such methods are often incomplete and lead to unnatural predictions, especially for portraits with hard shadows. We overcome the limitations of existing local propagation methods by formulating the removal problem as a generation task where a diffusion model learns to globally rebuild the human appearance from scratch as a condition of an input portrait image. For robust and natural shadow removal, we propose to train the diffusion model with a compositional repurposing framework: a pre-trained text-guided image generation model is first fine-tuned to harmonize the lighting and color of the foreground with a background scene by using a background harmonization dataset; and then the model is further fine-tuned to generate a shadow-free portrait image via a shadow-paired dataset. To overcome the limitation of losing fine details in the latent diffusion model, we propose a guided-upsampling network to restore the original high-frequency details (e.g. , wrinkles and dots) from the input image. To enable our compositional training framework, we construct a high-fidelity and large-scale dataset using a lightstage capturing system and synthetic graphics simulation. Our generative framework effectively removes shadows caused by both self and external occlusions while maintaining original lighting distribution and high-frequency details. Our method also demonstrates robustness to diverse subjects captured in real environments.