Vector Quantized Variational Autoencoders (VQ-VAEs) are fundamental models that compress continuous visual data into discrete tokens. Existing methods have tried to improve the quantization strategy for better reconstruction quality, however, there still exists a large gap between VQ-VAEs and VAEs. To narrow this gap, we propose MGVQ, a novel method to augment the representation capability of discrete codebooks, facilitating easier optimization for codebooks and minimizing information loss, thereby enhancing reconstruction quality. Specifically, we propose to retain the latent dimension to preserve encoded features and incorporate a set of sub-codebooks for quantization. Furthermore, we construct comprehensive zero-shot benchmarks featuring resolutions of 512p and 2k to evaluate the reconstruction performance of existing methods rigorously. MGVQ achieves the state-of-the-art performance on both ImageNet and 8 zero-shot benchmarks across all VQ-VAEs. Notably, compared with SD-VAE, we outperform them on ImageNet significantly, with rFID 0.49 v.s. 0.91, and achieve superior PSNR on all zero-shot benchmarks. These results highlight the superiority of MGVQ in reconstruction and pave the way for preserving fidelity in HD image processing tasks. Code will be publicly available at https://github.com/MKJia/MGVQ
Low-dose computed tomography (LDCT) reduces radiation exposure but often degrades image quality, potentially compromising diagnostic accuracy. Existing deep learning-based denoising methods focus primarily on pixel-level mappings, overlooking the potential benefits of high-level semantic guidance. Recent advances in vision-language models (VLMs) suggest that language can serve as a powerful tool for capturing structured semantic information, offering new opportunities to improve LDCT reconstruction. In this paper, we introduce LangMamba, a Language-driven Mamba framework for LDCT denoising that leverages VLM-derived representations to enhance supervision from normal-dose CT (NDCT). LangMamba follows a two-stage learning strategy. First, we pre-train a Language-guided AutoEncoder (LangAE) that leverages frozen VLMs to map NDCT images into a semantic space enriched with anatomical information. Second, we synergize LangAE with two key components to guide LDCT denoising: Semantic-Enhanced Efficient Denoiser (SEED), which enhances NDCT-relevant local semantic while capturing global features with efficient Mamba mechanism, and Language-engaged Dual-space Alignment (LangDA) Loss, which ensures that denoised images align with NDCT in both perceptual and semantic spaces. Extensive experiments on two public datasets demonstrate that LangMamba outperforms conventional state-of-the-art methods, significantly improving detail preservation and visual fidelity. Remarkably, LangAE exhibits strong generalizability to unseen datasets, thereby reducing training costs. Furthermore, LangDA loss improves explainability by integrating language-guided insights into image reconstruction and offers a plug-and-play fashion. Our findings shed new light on the potential of language as a supervisory signal to advance LDCT denoising. The code is publicly available on https://github.com/hao1635/LangMamba.
Cloud contamination significantly impairs the usability of optical satellite imagery, affecting critical applications such as environmental monitoring, disaster response, and land-use analysis. This research presents a Cloud-Attentive Reconstruction Framework that integrates SAR-optical feature fusion with deep learning-based image reconstruction to generate cloud-free optical imagery. The proposed framework employs an attention-driven feature fusion mechanism to align complementary structural information from Synthetic Aperture Radar (SAR) with spectral characteristics from optical data. Furthermore, a cloud-aware model update strategy introduces adaptive loss weighting to prioritize cloud-occluded regions, enhancing reconstruction accuracy. Experimental results demonstrate that the proposed method outperforms existing approaches, achieving a PSNR of 31.01 dB, SSIM of 0.918, and MAE of 0.017. These outcomes highlight the framework's effectiveness in producing high-fidelity, spatially and spectrally consistent cloud-free optical images.
Deep neural networks have been shown to achieve exceptional performance for computer vision tasks like image recognition, segmentation, and reconstruction or denoising. Here, we evaluate the ultimate performance limits of deep convolutional neural network models for image reconstruction, by comparing them against the standard quantum limit set by shot-noise and the Heisenberg limit on precision. We train U-Net models on images of natural objects illuminated with coherent states of light, and find that the average mean-squared error of the reconstructions can surpass the standard quantum limit, and in some cases reaches the Heisenberg limit. Further, we train models on well-parameterized images for which we can calculate the quantum Cram\'er-Rao bound to determine the minimum possible measurable variance of an estimated parameter for a given probe state. We find the mean-squared error of the model predictions reaches these bounds calculated for the parameters, across a variety of parameterized images. These results suggest that deep convolutional neural networks can learn to become the optimal estimators allowed by the laws of physics, performing parameter estimation and image reconstruction at the ultimate possible limits of precision for the case of classical illumination of the object.
We study the challenging problem of unsupervised multi-object segmentation on single images. Existing methods, which rely on image reconstruction objectives to learn objectness or leverage pretrained image features to group similar pixels, often succeed only in segmenting simple synthetic objects or discovering a limited number of real-world objects. In this paper, we introduce unMORE, a novel two-stage pipeline designed to identify many complex objects in real-world images. The key to our approach involves explicitly learning three levels of carefully defined object-centric representations in the first stage. Subsequently, our multi-object reasoning module utilizes these learned object priors to discover multiple objects in the second stage. Notably, this reasoning module is entirely network-free and does not require human labels. Extensive experiments demonstrate that unMORE significantly outperforms all existing unsupervised methods across 6 real-world benchmark datasets, including the challenging COCO dataset, achieving state-of-the-art object segmentation results. Remarkably, our method excels in crowded images where all baselines collapse.
CLIP is a discriminative model trained to align images and text in a shared embedding space. Due to its multimodal structure, it serves as the backbone of many generative pipelines, where a decoder is trained to map from the shared space back to images. In this work, we show that image synthesis is nevertheless possible using CLIP alone -- without any decoder, training, or fine-tuning. Our approach optimizes a frequency-aware implicit neural representation that encourages coarse-to-fine generation by stratifying frequencies across network layers. To stabilize this inverse mapping, we introduce adversarially robust initialization, a lightweight Orthogonal Procrustes projection to align local text and image embeddings, and a blending loss that anchors outputs to natural image statistics. Without altering CLIP's weights, this framework unlocks capabilities such as text-to-image generation, style transfer, and image reconstruction. These findings suggest that discriminative models may hold untapped generative potential, hidden in plain sight.
We propose a self-supervised feature learning assisted reconstruction (SSFL-Recon) framework for MRI reconstruction to address the limitation of existing supervised learning methods. Although recent deep learning-based methods have shown promising performance in MRI reconstruction, most require fully-sampled images for supervised learning, which is challenging in practice considering long acquisition times under respiratory or organ motion. Moreover, nearly all fully-sampled datasets are obtained from conventional reconstruction of mildly accelerated datasets, thus potentially biasing the achievable performance. The numerous undersampled datasets with different accelerations in clinical practice, hence, remain underutilized. To address these issues, we first train a self-supervised feature extractor on undersampled images to learn sampling-insensitive features. The pre-learned features are subsequently embedded in the self-supervised reconstruction network to assist in removing artifacts. Experiments were conducted retrospectively on an in-house 2D cardiac Cine dataset, including 91 cardiovascular patients and 38 healthy subjects. The results demonstrate that the proposed SSFL-Recon framework outperforms existing self-supervised MRI reconstruction methods and even exhibits comparable or better performance to supervised learning up to $16\times$ retrospective undersampling. The feature learning strategy can effectively extract global representations, which have proven beneficial in removing artifacts and increasing generalization ability during reconstruction.
DeepInverse is an open-source PyTorch-based library for solving imaging inverse problems. The library covers all crucial steps in image reconstruction from the efficient implementation of forward operators (e.g., optics, MRI, tomography), to the definition and resolution of variational problems and the design and training of advanced neural network architectures. In this paper, we describe the main functionality of the library and discuss the main design choices.
Advanced autoregressive models have reshaped multimodal AI. However, their transformative potential in medical imaging remains largely untapped due to the absence of a unified visual tokenizer -- one capable of capturing fine-grained visual structures for faithful image reconstruction and realistic image synthesis, as well as rich semantics for accurate diagnosis and image interpretation. To this end, we present MedITok, the first unified tokenizer tailored for medical images, encoding both low-level structural details and high-level clinical semantics within a unified latent space. To balance these competing objectives, we introduce a novel two-stage training framework: a visual representation alignment stage that cold-starts the tokenizer reconstruction learning with a visual semantic constraint, followed by a textual semantic representation alignment stage that infuses detailed clinical semantics into the latent space. Trained on the meticulously collected large-scale dataset with over 30 million medical images and 2 million image-caption pairs, MedITok achieves state-of-the-art performance on more than 30 datasets across 9 imaging modalities and 4 different tasks. By providing a unified token space for autoregressive modeling, MedITok supports a wide range of tasks in clinical diagnostics and generative healthcare applications. Model and code will be made publicly available at: https://github.com/Masaaki-75/meditok.
The use of denoisers for image reconstruction has shown significant potential, especially for the Plug-and-Play (PnP) framework. In PnP, a powerful denoiser is used as an implicit regularizer in proximal algorithms such as ISTA and ADMM. The focus of this work is on the convergence of PnP iterates for linear inverse problems using kernel denoisers. It was shown in prior work that the update operator in standard PnP is contractive for symmetric kernel denoisers under appropriate conditions on the denoiser and the linear forward operator. Consequently, we could establish global linear convergence of the iterates using the contraction mapping theorem. In this work, we develop a unified framework to establish global linear convergence for symmetric and nonsymmetric kernel denoisers. Additionally, we derive quantitative bounds on the contraction factor (convergence rate) for inpainting, deblurring, and superresolution. We present numerical results to validate our theoretical findings.
Brain-to-image decoding has been recently propelled by the progress in generative AI models and the availability of large ultra-high field functional Magnetic Resonance Imaging (fMRI). However, current approaches depend on complicated multi-stage pipelines and preprocessing steps that typically collapse the temporal dimension of brain recordings, thereby limiting time-resolved brain decoders. Here, we introduce Dynadiff (Dynamic Neural Activity Diffusion for Image Reconstruction), a new single-stage diffusion model designed for reconstructing images from dynamically evolving fMRI recordings. Our approach offers three main contributions. First, Dynadiff simplifies training as compared to existing approaches. Second, our model outperforms state-of-the-art models on time-resolved fMRI signals, especially on high-level semantic image reconstruction metrics, while remaining competitive on preprocessed fMRI data that collapse time. Third, this approach allows a precise characterization of the evolution of image representations in brain activity. Overall, this work lays the foundation for time-resolved brain-to-image decoding.
Autoregressive (AR) models have recently shown strong performance in image generation, where a critical component is the visual tokenizer (VT) that maps continuous pixel inputs to discrete token sequences. The quality of the VT largely defines the upper bound of AR model performance. However, current discrete VTs fall significantly behind continuous variational autoencoders (VAEs), leading to degraded image reconstructions and poor preservation of details and text. Existing benchmarks focus on end-to-end generation quality, without isolating VT performance. To address this gap, we introduce VTBench, a comprehensive benchmark that systematically evaluates VTs across three core tasks: Image Reconstruction, Detail Preservation, and Text Preservation, and covers a diverse range of evaluation scenarios. We systematically assess state-of-the-art VTs using a set of metrics to evaluate the quality of reconstructed images. Our findings reveal that continuous VAEs produce superior visual representations compared to discrete VTs, particularly in retaining spatial structure and semantic detail. In contrast, the degraded representations produced by discrete VTs often lead to distorted reconstructions, loss of fine-grained textures, and failures in preserving text and object integrity. Furthermore, we conduct experiments on GPT-4o image generation and discuss its potential AR nature, offering new insights into the role of visual tokenization. We release our benchmark and codebase publicly to support further research and call on the community to develop strong, general-purpose open-source VTs.
Recent advances in deep learning-based parallel compressed sensing magnetic resonance imaging (p-CSMRI) have significantly improved reconstruction quality. However, current p-CSMRI methods often require training separate deep neural network (DNN) for each organ due to anatomical variations, creating a barrier to developing generalized medical image reconstruction systems. To address this, we propose CAPNet (cross-organ all-in-one deep unfolding p-CSMRI network), a unified framework that implements a p-CSMRI iterative algorithm via three specialized modules: auxiliary variable module, prior module, and data consistency module. Recognizing that p-CSMRI systems often employ varying sampling ratios for different organs, resulting in organ-specific artifact patterns, we introduce an artifact generation submodule, which extracts and integrates artifact features into the data consistency module to enhance the discriminative capability of the overall network. For the prior module, we design an organ structure-prompt generation submodule that leverages structural features extracted from the segment anything model (SAM) to create cross-organ prompts. These prompts are strategically incorporated into the prior module through an organ structure-aware Mamba submodule. Comprehensive evaluations on a cross-organ dataset confirm that CAPNet achieves state-of-the-art reconstruction performance across multiple anatomical structures using a single unified model. Our code will be published at https://github.com/shibaoshun/CAPNet.
Recent advances in representation learning often rely on holistic, black-box embeddings that entangle multiple semantic components, limiting interpretability and generalization. These issues are especially critical in medical imaging. To address these limitations, we propose an Organ-Wise Tokenization (OWT) framework with a Token Group-based Reconstruction (TGR) training paradigm. Unlike conventional approaches that produce holistic features, OWT explicitly disentangles an image into separable token groups, each corresponding to a distinct organ or semantic entity. Our design ensures each token group encapsulates organ-specific information, boosting interpretability, generalization, and efficiency while allowing fine-grained control in downstream tasks. Experiments on CT and MRI datasets demonstrate the effectiveness of OWT in not only achieving strong image reconstruction and segmentation performance, but also enabling novel semantic-level generation and retrieval applications that are out of reach for standard holistic embedding methods. These findings underscore the potential of OWT as a foundational framework for semantically disentangled representation learning, offering broad scalability and applicability to real-world medical imaging scenarios and beyond.
Visual text is a crucial component in both document and scene images, conveying rich semantic information and attracting significant attention in the computer vision community. Beyond traditional tasks such as text detection and recognition, visual text processing has witnessed rapid advancements driven by the emergence of foundation models, including text image reconstruction and text image manipulation. Despite significant progress, challenges remain due to the unique properties that differentiate text from general objects. Effectively capturing and leveraging these distinct textual characteristics is essential for developing robust visual text processing models. In this survey, we present a comprehensive, multi-perspective analysis of recent advancements in visual text processing, focusing on two key questions: (1) What textual features are most suitable for different visual text processing tasks? (2) How can these distinctive text features be effectively incorporated into processing frameworks? Furthermore, we introduce VTPBench, a new benchmark that encompasses a broad range of visual text processing datasets. Leveraging the advanced visual quality assessment capabilities of multimodal large language models (MLLMs), we propose VTPScore, a novel evaluation metric designed to ensure fair and reliable evaluation. Our empirical study with more than 20 specific models reveals substantial room for improvement in the current techniques. Our aim is to establish this work as a fundamental resource that fosters future exploration and innovation in the dynamic field of visual text processing. The relevant repository is available at https://github.com/shuyansy/Visual-Text-Processing-survey.
Masked Image Modeling (MIM) with Vector Quantization (VQ) has achieved great success in both self-supervised pre-training and image generation. However, most existing methods struggle to address the trade-off in shared latent space for generation quality vs. representation learning and efficiency. To push the limits of this paradigm, we propose MergeVQ, which incorporates token merging techniques into VQ-based generative models to bridge the gap between image generation and visual representation learning in a unified architecture. During pre-training, MergeVQ decouples top-k semantics from latent space with the token merge module after self-attention blocks in the encoder for subsequent Look-up Free Quantization (LFQ) and global alignment and recovers their fine-grained details through cross-attention in the decoder for reconstruction. As for the second-stage generation, we introduce MergeAR, which performs KV Cache compression for efficient raster-order prediction. Extensive experiments on ImageNet verify that MergeVQ as an AR generative model achieves competitive performance in both visual representation learning and image generation tasks while maintaining favorable token efficiency and inference speed. The code and model will be available at https://apexgen-x.github.io/MergeVQ.
In autoregressive (AR) image generation, visual tokenizers compress images into compact discrete latent tokens, enabling efficient training of downstream autoregressive models for visual generation via next-token prediction. While scaling visual tokenizers improves image reconstruction quality, it often degrades downstream generation quality -- a challenge not adequately addressed in existing literature. To address this, we introduce GigaTok, the first approach to simultaneously improve image reconstruction, generation, and representation learning when scaling visual tokenizers. We identify the growing complexity of latent space as the key factor behind the reconstruction vs. generation dilemma. To mitigate this, we propose semantic regularization, which aligns tokenizer features with semantically consistent features from a pre-trained visual encoder. This constraint prevents excessive latent space complexity during scaling, yielding consistent improvements in both reconstruction and downstream autoregressive generation. Building on semantic regularization, we explore three key practices for scaling tokenizers:(1) using 1D tokenizers for better scalability, (2) prioritizing decoder scaling when expanding both encoder and decoder, and (3) employing entropy loss to stabilize training for billion-scale tokenizers. By scaling to $\bf{3 \space billion}$ parameters, GigaTok achieves state-of-the-art performance in reconstruction, downstream AR generation, and downstream AR representation quality.
Gaussian mixture models (GMMs) are widely used in machine learning for tasks such as clustering, classification, image reconstruction, and generative modeling. A key challenge in working with GMMs is defining a computationally efficient and geometrically meaningful metric. The mixture Wasserstein (MW) distance adapts the Wasserstein metric to GMMs and has been applied in various domains, including domain adaptation, dataset comparison, and reinforcement learning. However, its high computational cost -- arising from repeated Wasserstein distance computations involving matrix square root estimations and an expensive linear program -- limits its scalability to high-dimensional and large-scale problems. To address this, we propose multiple novel slicing-based approximations to the MW distance that significantly reduce computational complexity while preserving key optimal transport properties. From a theoretical viewpoint, we establish several weak and strong equivalences between the introduced metrics, and show the relations to the original MW distance and the well-established sliced Wasserstein distance. Furthermore, we validate the effectiveness of our approach through numerical experiments, demonstrating computational efficiency and applications in clustering, perceptual image comparison, and GMM minimization
Learning-based image compression methods have recently emerged as promising alternatives to traditional codecs, offering improved rate-distortion performance and perceptual quality. JPEG AI represents the latest standardized framework in this domain, leveraging deep neural networks for high-fidelity image reconstruction. In this study, we present a comprehensive subjective visual quality assessment of JPEG AI-compressed images using the JPEG AIC-3 methodology, which quantifies perceptual differences in terms of Just Noticeable Difference (JND) units. We generated a dataset of 50 compressed images with fine-grained distortion levels from five diverse sources. A large-scale crowdsourced experiment collected 96,200 triplet responses from 459 participants. We reconstructed JND-based quality scales using a unified model based on boosted and plain triplet comparisons. Additionally, we evaluated the alignment of multiple objective image quality metrics with human perception in the high-fidelity range. The CVVDP metric achieved the overall highest performance; however, most metrics including CVVDP were overly optimistic in predicting the quality of JPEG AI-compressed images. These findings emphasize the necessity for rigorous subjective evaluations in the development and benchmarking of modern image codecs, particularly in the high-fidelity range. Another technical contribution is the introduction of the well-known Meng-Rosenthal-Rubin statistical test to the field of Quality of Experience research. This test can reliably assess the significance of difference in performance of quality metrics in terms of correlation between metrics and ground truth. The complete dataset, including all subjective scores, is publicly available at https://github.com/jpeg-aic/dataset-JPEG-AI-SDR25.
Multimodal misinformation, such as miscaptioned images, where captions misrepresent an image's origin, context, or meaning, poses a growing challenge in the digital age. To support fact-checkers, researchers have been focusing on creating datasets and developing methods for multimodal misinformation detection (MMD). Due to the scarcity of large-scale annotated MMD datasets, recent studies leverage synthetic training data via out-of-context image-caption pairs or named entity manipulations; altering names, dates, and locations. However, these approaches often produce simplistic misinformation that fails to reflect real-world complexity, limiting the robustness of detection models trained on them. Meanwhile, despite recent advancements, Large Vision-Language Models (LVLMs) remain underutilized for generating diverse, realistic synthetic training data for MMD. To address this gap, we introduce "MisCaption This!", a training dataset comprising LVLM-generated miscaptioned images. Additionally, we introduce "Latent Multimodal Reconstruction" (LAMAR), a network trained to reconstruct the embeddings of truthful captions, providing a strong auxiliary signal to the detection process. To optimize LAMAR, we explore different training strategies (end-to-end training and large-scale pre-training) and integration approaches (direct, mask, gate, and attention). Extensive experiments show that models trained on "MisCaption This!" generalize better on real-world misinformation, while LAMAR sets new state-of-the-art on both NewsCLIPpings and VERITE benchmarks; highlighting the potential of LVLM-generated data and reconstruction-based approaches for advancing MMD. We release our code at: https://github.com/stevejpapad/miscaptioned-image-reconstruction