Search Results (633)

Image super-resolution (SR) is a fundamental task in computer vision that aims to recover high-fidelity details from low-resolution images. To address the limitations of existing Transformer-based SR methods, this paper introduces GSDAT-Net, a grid-spatial dual-attention-driven residual hybrid Transformer. By integrating Grid Attention Block (GAB), an Enhanced Spatial Attention (ESA) module, and SwinV2 Transformer layers (S2TL), GSDAT-Net extracts multi-scale features and improves local and global feature representation. Extensive experiments demonstrate that GSDAT-Net achieves competitive reconstruction accuracy and visual quality compared with the selected methods, with a maximum PSNR improvement of up to 0.15 dB on benchmark datasets. Full article

The super-resolution (SR) of Magnetic Resonance Imaging (MRI) is essential for utilizing clinical scans with limited resolution, noise, and anisotropic sampling, such as multi-modal brain tumor imaging. In this work, we propose a Tri-Representation hybrid framework for MRI SR, TRT-GLA, that redefines the MRI SR task as a joint spatial–spectral–structural high-resolution image generation problem. TRT-GLA utilizes (i) spatial global–local attentions for modeling the spatial anatomy, (ii) a Fourier spectral transfer mechanism for upholding spectral consistency, and (iii) multi-scale hierarchical spectral decomposition for improved edge details. To adapt the learning framework to medical imaging characteristics, we introduce a tri-representation consistent loss function that explicitly combines pixel-wise, spectral, and edge structure priors from the high-resolution ground-truth, as well as a progressive resolution learning strategy. Our large-scale brain tumor experiments, on the IXI, BraTS 2019, 2020, and 2023 datasets, show that TRT-GLA achieves state-of-the-art results at upsampling factors of ×2, ×4, and ×8, respectively, achieving substantial improvements across CNN, GAN, and transformer-based methods in Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Multi-scale Structural Similarity Index (MS-SSIM). We further demonstrate how SR benefits brain tumor segmentation through the downstream task evaluation of a dual-branch segmentation framework. TRT-GLA produces highly accurate tumor segmentation results from low-resolution inputs, improving over native high-resolution inputs at ×8 in critical tumor boundary regions and in small tumor regions. There remains a small gap between native, high-resolution imaging and SR-enhanced performance, which TRT-GLA nearly closes under realistic scenarios. Our results highlight the importance of synthesizing unified priors over spatial, spectral, and structural domains within a transformer for anatomically faithful reconstructions. Importantly, we also establish the utility of TRT-GLA in supporting quantitative analysis through a downstream tumor segmentation experiment that is clinically relevant. Full article

High dynamic range (HDR) radiance-field reconstruction from single-exposure low dynamic range (LDR) images is limited by the information loss in saturated regions, while event cameras provide complementary measurements whose dynamic range far exceeds that of conventional sensors. In practical hybrid sensor systems, however, the RGB camera usually has a higher spatial resolution than the event sensor, which makes existing event-aided HDR reconstruction methods difficult to apply directly. A straightforward solution is to super-resolve the event stream before reconstruction, but this 2D preprocessing introduces a global color cast and view-inconsistent high-frequency artifacts once the super-resolved events supervise a 3D radiance field. We propose a framework for HDR radiance-field reconstruction from resolution-mismatched event–image inputs. The framework incorporates a pretrained 2D event super-resolution prior and corrects its transfer to 3D reconstruction through a color correction module, which anchors the rendered radiance to the chrominance of the LDR images, and a dual-resolution event-stream constraint, which supervises the synthesized events at both the super-resolved and the native resolution. Experiments on the EvHDR-NeRF benchmark show that the proposed method achieves higher six-scene mean HDR fidelity than the strongest baseline while reducing the global color cast and alleviating multi-view artifacts. Full article

Text image super-resolution aims to improve the readability of low-quality text images while preserving character structures, stroke details, and semantic consistency. Compared with natural image super-resolution, this task is more sensitive to structural distortion because small changes in stroke topology may lead to incorrect text recognition. To address this problem, this paper proposes an OCR prior-guided cross-scale framework for text image super-resolution. Specifically, character-level semantic priors extracted from a pretrained OCR model are introduced to provide structural guidance for degraded text reconstruction. A gated feature modulation mechanism is designed to adaptively regulate the contribution of OCR priors, reducing the influence of unreliable semantic predictions. A cross-scale dynamic attention module is also developed to aggregate multi-granularity visual features, enabling the model to jointly recover fine stroke boundaries and global character structures. In addition, a sequence-aware calibration module is introduced to improve structural consistency along the logical reading order of text. Experiments on mixed text image benchmarks and the TextZoom dataset show that the proposed method achieves competitive or better performance among the compared methods in terms of PSNR, SSIM, and recognition-oriented metrics. Additional ablation, OCR prior robustness, and computational complexity analyses further indicate that the proposed framework improves text readability while maintaining a reasonable accuracy–complexity trade-off. The results also suggest that OCR priors are useful for text image reconstruction, but should be used as soft constraints when external recognition predictions are uncertain. Full article

Existing image super-resolution methods still suffer from limitations in edge-structure restoration, high-frequency texture preservation, and artifact suppression, which may lead to blurred contours and unnatural textures. To address these issues, this paper proposes an image super-resolution method based on hierarchical feature aggregation and Laplacian high-frequency compensation. First, a Hierarchical Feature Aggregation Attention Block (HFAB) is designed in the generator to progressively extract image features at different levels through multiple convolutional layers. A High-Frequency Variance Adaptive Channel Attention Block (HFVB) is further introduced to adaptively enhance key texture and edge information. Second, a Laplacian Adaptive Upsampling (LAU) module is developed to combine low-frequency content reconstruction with high-frequency detail compensation, thereby strengthening edge contours, preserving fine textures, and reducing artifacts. Finally, a Dissimilarity Structural Similarity Index Measure (DSSIM) loss is incorporated into the loss function to constrain local structural consistency and further improve the structural preservation and perceptual quality of reconstructed images. Experimental results on Set5, Set14, BSD100, and Urban100 show that, compared with SRGAN, the proposed method improves PSNR by 0.31 dB, 0.18 dB, 0.14 dB, and 0.14 dB, respectively, while reducing LPIPS by 0.0234, 0.0183, 0.0222, and 0.0222. These results indicate that the proposed method provides consistent improvements over SRGAN and achieves modest, metric-dependent gains over ESRGAN, suggesting an incremental enhancement in reconstruction accuracy and perceptual quality on both natural image benchmarks and complex urban scene datasets. Full article

Conventional imaging systems often suffer from a coupled limitation of narrow field of view, single modality, and insufficient effective resolution. Wide-angle imaging preserves scene context but compresses distant or small-scale targets into limited pixels, while narrow-field imaging improves details at the cost of global perception. Moreover, single-modal visible imaging is sensitive to illumination and contrast variations, whereas infrared imaging lacks fine spatial texture. To increase information at the imaging source, we propose coupled spectral–spatial fusion-enabled multi-scale panoramic imaging, a dual-field-of-view (FOV) visible–infrared framework for wide-field high-resolution perception. Two imaging units acquire paired visible and infrared images from adjacent overlapping views. For each view, a visible–infrared fusion super-resolution model integrates visible structural details with infrared radiative cues to reconstruct a high-resolution fused image. A multi-scale stitching algorithm then extracts robust features, estimates cross-view correspondences, and merges the two fused images into a large-FOV panoramic result. Outdoor experiments demonstrate that the proposed method improves local contrast, suppresses pixelation artifacts, enhances readable fine details, and expands the observable field of view, providing an effective route toward multimodal panoramic imaging. Full article

Diffusion-based image super-resolution (SR) has recently achieved impressive perceptual quality by progressively generating plausible high-resolution details. However, its restoration performance still depends strongly on the reliability of the conditioning signal derived from degraded low-resolution inputs. Under severe or complex degradations, LR-derived conditions may become incomplete or ambiguous, leading the denoising trajectory toward visually plausible but input-inconsistent reconstructions. This work focuses on a central question: how to construct reliable multimodal prior guidance for a diffusion backbone that commonly adopts a hierarchical U-shaped architecture. To this end, we propose STMP-DiT, a stage-aware text-aligned multimodal prior-guided Diffusion Transformer for image super-resolution. From a multimodal data mining perspective, STMP-DiT aims to discover, align, and organize complementary semantic and structural priors from heterogeneous foundation-model representations. To improve the reliability of semantic guidance, STMP-DiT first aligns LLaVA-derived LR prompts with the frozen CLIP HR-image embedding space, producing visually grounded textual priors for restoration. These aligned textual priors are complemented by hierarchical DINO features, where deep features guide coarse semantic layout, intermediate features support structural recovery, and shallow features refine local edges and textures in the U-shaped DiT backbone. Rather than treating textual and visual priors as a single homogeneous condition, STMP-DiT assigns hierarchical DINO priors to different restoration stages according to their representational granularity. The fused condition is then injected through bounded feature modulation, enabling controlled stage-aware guidance while reducing redundant conditioning and improving the parameter efficiency of the conditioning modules. Experimental results on widely used SR benchmarks indicate promising improvements in perceptual quality and distributional realism with a compact trainable parameter scale, suggesting the value of reliable, stage-aware, and parameter-efficient multimodal prior integration for diffusion-based image super-resolution. Full article

Recent advances in generative artificial intelligence (AI) have shown significant promise for brain magnetic resonance imaging (MRI), enabling applications such as image synthesis, modality translation, reconstruction, super-resolution, segmentation, anomaly detection, and disease identification. This PRISMA-ScR-guided scoping review provides a structured synthesis of recent peer-reviewed studies on generative AI for brain MRI analysis published between January 2024 and March 2026. A total of 43 studies meeting predefined inclusion criteria were analyzed. We review major generative architectures, including variational autoencoders (VAEs), generative adversarial networks (GANs), diffusion models, and transformer-based generative models, and summarize their applications across key neuroimaging tasks. We also provide an overview of the publicly available datasets commonly used for model development and evaluation. Beyond reporting performance, this review critically examines the current evidence with respect to reproducibility, external validation, data availability, evaluation validity, data leakage, hallucination and safety risks, and barriers to clinical translation. Although many studies report promising results on retrospective benchmark datasets, external validation, prospective evaluation, reader studies, and clinically oriented assessments remain relatively uncommon. Challenges related to generalization, dataset heterogeneity, computational requirements, privacy, and regulatory considerations continue to limit real-world deployment. Overall, the reviewed literature demonstrates that generative AI has substantial potential to improve brain MRI analysis through realistic data generation, enhanced image quality, and more informative feature representations. However, the current evidence primarily supports technical feasibility and methodological advances rather than established clinical utility. We conclude by identifying key research gaps and future research directions toward more robust, interpretable, reproducible, and clinically translatable generative AI frameworks for brain MRI analysis. Full article

Remote sensing imagery provides a primary data source for large-scale surface water body monitoring, which is crucial for quantifying climate-related hydrological impacts, supporting flood control, and sustaining integrated water resource management. However, remote sensing images generally face the trade-off between spatial resolution and temporal coverage. To address this issue, the Real-ESRGAN super-resolution algorithm is employed to reconstruct temporally continuous, wide-coverage medium-resolution imagery to a 2.5 m resolution, effectively improving its capability to identify sub-pixel river boundaries. Water body segmentation (WBS) is an effective method for fine-detail surface water extraction. Nonetheless, when applied in complex hydrological environments, it still faces several limitations, such as ambiguous delineation of land–water boundaries and the difficulty in capturing multiscale water body characteristics. To address these issues, a Dynamic Weight Fusion SegFormer (DWFSeg) network is constructed, integrating a MixVision Transformer (MVT) encoder with a multiscale decoding architecture. Specifically, a Dynamic Multiscale Feature Fusion (DMFF) mechanism is proposed, which adaptively assigns semantic-guided fusion weights to multiscale feature water bodies. Furthermore, the Dual Attention-Enhanced (DAE) module strengthens discriminative essential features and suppresses background noise in both channel and spatial dimensions. Evaluated on a self-constructed super-resolution imagery dataset (SID) and the public GID, DWFSeg achieves overall accuracies of 98.08% and 96.14%, respectively. It outperforms representative benchmark models across multiple quantitative metrics, while maintaining competitive inference efficiency and favorable segmentation stability. Ablation studies verify the effectiveness and necessity of each proposed component. The presented network provides a reliable technical solution and supports refined water resource evaluation and sustainable watershed management. Full article

Remote sensing image super-resolution plays an important role in object recognition, urban monitoring, and fine-grained remote sensing interpretation. This paper studies lightweight single-image remote sensing image super-resolution, in which only one LR observation is available and the model must recover reliable structural details under a limited computational budget. Existing lightweight methods reduce parameter counts and computational complexity, but their limited representation capacity often causes blurred boundaries, broken road structures, and missing high-frequency details in buildings, roads, and texture-rich regions. To address these issues, we propose SFEFeNet, a Structure-Frequency Mutual-Guided Lightweight Network for remote sensing image super-resolution. First, we design a Lightweight Structure-Frequency Block (LSFB) to jointly model local spatial features, structural responses, and frequency responses with low computational overhead. Second, we introduce a Structure-Frequency Mutual Guidance (SFMG) module, where edge responses guide high-frequency component selection, and the selected high-frequency responses further refine edge-aware attention. Finally, we propose a Structure-Frequency Fusion Gate (SFFG) to adaptively integrate lightweight features, local spatial features, frequency-enhanced features, and structure-refined features. Experiments on RSSCN7, DOTA, and WHU-RS19 datasets evaluate SFEFeNet in terms of reconstruction quality, visual performance, and model complexity. Additional analyses further examine structural preservation, complex synthetic degradation, real-image generalization, and statistical stability. Notably, SFEFeNet-Lite contains 0.539 M parameters and 17.07 G FLOPs for $\times 2$, and 0.622 M parameters and 7.12 G FLOPs for $\times 4$, enabling effective structure-frequency feature modeling with lightweight computational cost. Full article

We propose a dual-aperture, simultaneous-polarization super-resolution imaging system that combines a total internal reflection optical architecture with a digital micromirror device (DMD) for broadband, high-resolution imaging. The system captures multiple polarization states simultaneously with a single detector and offers a compact, lightweight design. Using reflective Wassermann–Wolf differential equations and Seidel aberration theory, we establish astigmatism-correction boundary conditions and apply iterative optimization to jointly correct spherical aberration, coma, astigmatism, and distortion. Because distortion critically affects super-resolution reconstruction by causing mirror–pixel misregistration, we further introduce a custom merit function to tightly constrain chief-ray positions for each sub-aperture and field point on intermediate and final image planes, effectively suppressing distortion. The final design achieves F/2.5, grid distortion below ±0.5%, and near-diffraction-limited performance in all polarization channels. Tolerance analysis of the four sub-apertures confirms that imaging requirements are satisfied, demonstrating robust high-resolution polarization imaging across multiple polarization states. Full article

In airborne sensing applications, radar forward-looking imaging is a crucial technology for high-resolution ground mapping and terrain perception. Super-resolution deconvolution is key to overcoming the real-beam resolution limits of these airborne sensors. However, when utilizing phased array scanning radars for wide-swath ground mapping, the antenna pattern exhibits severe spatial variation at large scanning angles, which directly leads to model mismatch and degradation in super-resolution performance. To address this hardware-induced sensing limitation, this paper proposes a sparse super-resolution method tailored for forward-looking phased array scanning radar. Firstly, the causes of the spatial variation in antenna pattern are analyzed, and a modified antenna convolution matrix is derived to accurately model the scanning process. Secondly, the corresponding objective function is formulated under the assumption of target sparsity. Finally, an alternating direction method of multipliers (ADMM) solver based on reweighted strategy is employed to resolve the objective function. Experimental results demonstrate that the proposed method achieves approximately a 4 times increase in cross-range resolution and effectively enhances the observation capabilities within the radar forward-looking area. Full article

Face super-resolution (FSR) aims to reconstruct high-quality high-resolution face images from low-resolution inputs. Although CNN–Transformer hybrid models have shown promising performance by jointly modeling local textures and global dependencies, their large parameter sizes and high computational costs hinder practical deployment in resource-constrained scenarios such as mobile devices and embedded systems. Meanwhile, existing lightweight SR models usually reduce complexity by simplifying network depth, channel dimensions, or convolutional operations, which may weaken feature representation capability and lead to insufficient recovery of fine facial structures. To address these issues, this paper proposes HCTIUNet, a lightweight CNN–Transformer hybrid network based on an inverted U-shaped architecture. Specifically, the proposed network integrates lightweight CNN branches for local facial texture extraction and Transformer branches for global dependency modeling, while introducing a multi-scale feature interaction strategy and a global feature refinement module to enhance facial structural details. Experimental results on the FFHQ, CelebA, and Helen datasets demonstrate that HCTIUNet achieves competitive performance under the ×8 face super-resolution setting, obtaining PSNR/SSIM/LPIPS values of 27.55 dB/0.765/0.225, 27.63 dB/0.761/0.212, and 27.53 dB/0.777/0.213, respectively. Moreover, HCTIUNet contains 10.5 M parameters, requires 9.9 G FLOPs, and achieves an inference time of 0.021 s. These results indicate that the proposed method achieves a favorable trade-off between reconstruction accuracy, perceptual quality, and computational efficiency, making it suitable for efficient face super-resolution applications. Full article

In characterizing unconventional reservoirs, conventional Digital Rock Physics (DRP) has long been constrained by three fundamental bottlenecks: the trade-off between imaging resolution and field of view, challenges in reconstructing multiscale pore topology, and the prohibitive computational cost of direct numerical simulation (DNS) at the pore scale. The deep integration of artificial intelligence and rock physics has given rise to a new paradigm—Intelligent Digital Rock Physics (IDRP). This paper provides a systematic review of the evolutionary trajectory of IDRP, with a focus on how machine learning is reshaping the end-to-end workflow from imaging and segmentation to reconstruction and simulation. First, we survey image super-resolution and 3D pore structure generation techniques based on convolutional neural networks (CNNs), generative adversarial networks (GANs), and diffusion models, elucidating their mechanisms for surpassing optical diffraction limits and incorporating macroscopic petrophysical constraints. Second, we outline algorithmic strategies for fusing multi-source heterogeneous data (e.g., Micro-CT and SEM) and representing dual-porosity or multi-continuum systems. Third, we critically examine the application of machine learning surrogates in single- and multiphase flow prediction, highlighting how physics-informed machine learning (PIML) and reinforcement learning (RL)—by embedding governing equations such as Navier–Stokes or Muskat–Leverett into loss functions—achieve both computational acceleration and physical consistency. We further identify key limitations of current IDRP approaches, including insufficient validation of generated topological realism, narrow generalization across lithologies, inadequate representation of dynamic wettability, and limited model interpretability. Finally, we propose a forward-looking roadmap centered on multimodal foundation models for rocks, coupled with neural operators and uncertainty quantification frameworks, emphasizing the critical pathways for translating IDRP into engineering digital twins for unconventional hydrocarbon development, coalbed methane production enhancement, Enhanced Geothermal Systems, and geological CO₂ storage. This review offers a comprehensive reference for researchers at the intersection of geophysics, rock mechanics, and artificial intelligence. Full article

The Dense-Residual-Connected Transformer (DRCT) has established a new state-of-the-art in single image super-resolution by mitigating the information bottleneck in deep networks. However, its feature aggregation mechanism relies on a suboptimal Static Addition strategy, where residual features are scaled by a fixed, learnable scalar, regardless of the image content. This content-agnostic approach treats high-frequency textures and low-frequency noise indiscriminately, limiting the model’s representational capability. To address this, we propose a Frequency-Aware Adaptive Fusion Gate (FAFG) to replace the static scaling. Unlike spatial-only gating mechanisms, FAFG integrates the Discrete Cosine Transform (DCT) to explicitly perceive the frequency distribution of feature maps. By decomposing features into frequency components, our gate acts as an intelligent valve, dynamically amplifying valid structural details while suppressing redundant background noise. Extensive experiments on standard benchmarks demonstrate that our proposed FAFG-integrated model consistently outperforms the static-scaling and other state-of-the-art methods. Specifically, our method achieves a significant PSNR improvement of 0.31 dB on the texture-rich Urban100 dataset at $\times 4$ scale. Visual results further confirm that our frequency-aware gating mechanism effectively recovers sharper edges and fine textures, providing a superior trade-off between reconstruction accuracy and model complexity. Full article