Debabrata Mandal, Soumitri Chattopadhyay, Yujie Wang
et al.
Universal image restoration aims to recover clean images from arbitrary real-world degradations using a single inference model. Despite significant progress, existing all-in-one restoration networks do not scale to multiple degradations. As the number of degradations increases, training becomes unstable, models grow excessively large, and performance drops across both seen and unseen domains. In this work, we show that scaling universal restoration is fundamentally limited by interference across degradations during joint learning, leading to catastrophic task forgetting. To address this challenge, we introduce a unified inference pipeline with a multi-branch mixture-of-experts architecture that decomposes restoration knowledge across specialized task-adaptable experts. Our approach enables scalable learning (over sixteen degradations), adapts and generalizes robustly to unseen domains, and supports user-controllable restoration across degradations. Beyond achieving superior performance across benchmarks, this work establishes a new design paradigm for scalable and controllable universal image restoration.
Image restoration (IR) models are typically trained to recover high-quality images using L1 or LPIPS loss. To handle diverse unknown degradations, zero-shot IR methods have also been introduced. However, existing pre-trained and zero-shot IR approaches often fail to align with human preferences, resulting in restored images that may not be favored. This highlights the critical need to enhance restoration quality and adapt flexibly to various image restoration tasks or backbones without requiring model retraining and ideally without labor-intensive preference data collection. In this paper, we propose the first Test-Time Preference Optimization (TTPO) paradigm for image restoration, which enhances perceptual quality, generates preference data on-the-fly, and is compatible with any IR model backbone. Specifically, we design a training-free, three-stage pipeline: (i) generate candidate preference images online using diffusion inversion and denoising based on the initially restored image; (ii) select preferred and dispreferred images using automated preference-aligned metrics or human feedback; and (iii) use the selected preference images as reward signals to guide the diffusion denoising process, optimizing the restored image to better align with human preferences. Extensive experiments across various image restoration tasks and models demonstrate the effectiveness and flexibility of the proposed pipeline.
Twisha Chattopadhyay, Fabricio Ceschin, Marco E. Garza
et al.
The 3D printing industry is rapidly growing and increasingly adopted across various sectors including manufacturing, healthcare, and defense. However, the operational setup often involves hazardous environments, necessitating remote monitoring through cameras and other sensors, which opens the door to cyber-based attacks. In this paper, we show that an adversary with access to video recordings of the 3D printing process can reverse engineer the underlying 3D print instructions. Our model tracks the printer nozzle movements during the printing process and maps the corresponding trajectory into G-code instructions. Further, it identifies the correct parameters such as feed rate and extrusion rate, enabling successful intellectual property theft. To validate this, we design an equivalence checker that quantitatively compares two sets of 3D print instructions, evaluating their similarity in producing objects alike in shape, external appearance, and internal structure. Unlike simple distance-based metrics such as normalized mean square error, our equivalence checker is both rotationally and translationally invariant, accounting for shifts in the base position of the reverse engineered instructions caused by different camera positions. Our model achieves an average accuracy of 90.87 percent and generates 30.20 percent fewer instructions compared to existing methods, which often produce faulty or inaccurate prints. Finally, we demonstrate a fully functional counterfeit object generated by reverse engineering 3D print instructions from video.
Entanglement asymmetry -- used here as a direct probe of symmetry restoration -- provides a sharp diagnostic of post-quench dynamics. We test this idea in the complex Sachdev--Ye--Kitaev model with a conserved U(1) charge. Using exact diagonalization, we track the joint evolution of entanglement entropy and entanglement asymmetry after quenches from charge-asymmetric product states. We find rapid volume-law entanglement growth consistent with the subsystem eigenstate thermalization hypothesis, accompanied by a concurrent decay of entanglement asymmetry to a late-time plateau set by finite-size effects: small subsystems display near-complete restoration, while residual cross-sector weight yields a finite plateau. Notably, we uncover a quantum Mpemba effect: states prepared further from symmetry relax faster and approach lower residual asymmetry; disorder in the couplings renders this behavior more robust and monotonic across parameters. We further derive a Pinsker-type lower bound that ties the decay of asymmetry to differences in subsystem purity, identifying dephasing between U(1) charge sectors as the operative mechanism. These results establish entanglement asymmetry as a sensitive probe of symmetry restoration and thermalization, clarifying finite-size limits in fast-scrambling, closed quantum systems.
With the exponential increase in image data, training an image restoration model is laborious. Dataset distillation is a potential solution to this problem, yet current distillation techniques are a blank canvas in the field of image restoration. To fill this gap, we propose the Distribution-aware Dataset Distillation method (TripleD), a new framework that extends the principles of dataset distillation to image restoration. Specifically, TripleD uses a pre-trained vision Transformer to extract features from images for complexity evaluation, and the subset (the number of samples is much smaller than the original training set) is selected based on complexity. The selected subset is then fed through a lightweight CNN that fine-tunes the image distribution to align with the distribution of the original dataset at the feature level. To efficiently condense knowledge, the training is divided into two stages. Early stages focus on simpler, low-complexity samples to build foundational knowledge, while later stages select more complex and uncertain samples as the model matures. Our method achieves promising performance on multiple image restoration tasks, including multi-task image restoration, all-in-one image restoration, and ultra-high-definition image restoration tasks. Note that we can train a state-of-the-art image restoration model on an ultra-high-definition (4K resolution) dataset using only one consumer-grade GPU in less than 8 hours (500 savings in computing resources and immeasurable training time).
Jaewon Min, Jin Hyeon Kim, Paul Hyunbin Cho
et al.
Image restoration aims to recover degraded images. However, existing diffusion-based restoration methods, despite great success in natural image restoration, often struggle to faithfully reconstruct textual regions in degraded images. Those methods frequently generate plausible but incorrect text-like patterns, a phenomenon we refer to as text-image hallucination. In this paper, we introduce Text-Aware Image Restoration (TAIR), a novel restoration task that requires the simultaneous recovery of visual contents and textual fidelity. To tackle this task, we present SA-Text, a large-scale benchmark of 100K high-quality scene images densely annotated with diverse and complex text instances. Furthermore, we propose a multi-task diffusion framework, called TeReDiff, that integrates internal features from diffusion models into a text-spotting module, enabling both components to benefit from joint training. This allows for the extraction of rich text representations, which are utilized as prompts in subsequent denoising steps. Extensive experiments demonstrate that our approach consistently outperforms state-of-the-art restoration methods, achieving significant gains in text recognition accuracy. See our project page: https://cvlab-kaist.github.io/TAIR/
Freeform thin-shell surfaces are critical in various fields, but their fabrication is complex and costly. Traditional methods are wasteful and require custom molds, while 3D printing needs extensive support structures and post-processing. Thermoshrinkage actuated 4D printing is an effective method through flat structures fabricating 3D shell. However, existing research faces issues related to precise deformation and limited robustness. Addressing these issues is challenging due to three key factors: (1) Difficulty in finding a universal method to control deformation across different materials; (2) Variability in deformation influenced by factors such as printing speed, layer thickness, and heating temperature; (3) Environmental factors affecting the deformation process. To overcome these challenges, we introduce FreeShell, a robust 4D printing technique that uses thermoshrinkage to create precise 3D shells. This method prints triangular tiles connected by shrinkable connectors from a single material. Upon heating, the connectors shrink, moving the tiles to form the desired 3D shape, simplifying fabrication and reducing material and environment dependency. An optimized algorithm for flattening 3D meshes ensures precision in printing. FreeShell demonstrates its effectiveness through various examples and experiments, showcasing accuracy, robustness, and strength, representing advancement in fabricating complex freeform surfaces.
This work prioritizes building a modular pipeline that utilizes existing models to systematically restore images, rather than creating new restoration models from scratch. Restoration is carried out at an object-specific level, with each object regenerated using its corresponding class label information. The approach stands out by providing complete user control over the entire restoration process. Users can select models for specialized restoration steps, customize the sequence of steps to meet their needs, and refine the resulting regenerated image with depth awareness. The research provides two distinct pathways for implementing image regeneration, allowing for a comparison of their respective strengths and limitations. The most compelling aspect of this versatile system is its adaptability. This adaptability enables users to target particular object categories, including medical images, by providing models that are trained on those object classes.
The entanglement asymmetry is an observable independent tool to investigate the relaxation of quantum many body systems through the restoration of an initially broken symmetry of the dynamics. In this paper we use this to investigate the effects of interactions on quantum relaxation in a paradigmatic integrable model. Specifically, we study the dynamical restoration of the $U(1)$ symmetry corresponding to rotations about the $z$-axis in the XXZ model quenched from a tilted ferromagnetic state. We find two distinct patterns of behaviour depending upon the interaction regime of the model. In the gapless regime, at roots of unity, we find that the symmetry restoration is predominantly carried out by bound states of spinons of maximal length. The velocity of these bound states is suppressed as the anisotropy is decreased towards the isotropic point leading to slower symmetry restoration. By varying the initial tilt angle, one sees that symmetry restoration is slower for an initally smaller tilt angle, signifying the presence of the quantum Mpemba effect. In the gapped regime however, spin transport for non maximally tilted states, is dominated by smaller bound states with longer bound states becoming frozen. This leads to a much longer time scales for restoration compared to the gapless regime. In addition, the quantum Mpemba effect is absent in the gapped regime.
Image restoration, which aims to retrieve and enhance degraded images, is fundamental across a wide range of applications. While conventional deep learning approaches have notably improved the image quality across various tasks, they still suffer from (i) the high storage cost needed for various task-specific models and (ii) the lack of interactivity and flexibility, hindering their wider application. Drawing inspiration from the pronounced success of prompts in both linguistic and visual domains, we propose novel Prompt-In-Prompt learning for universal image restoration, named PIP. First, we present two novel prompts, a degradation-aware prompt to encode high-level degradation knowledge and a basic restoration prompt to provide essential low-level information. Second, we devise a novel prompt-to-prompt interaction module to fuse these two prompts into a universal restoration prompt. Third, we introduce a selective prompt-to-feature interaction module to modulate the degradation-related feature. By doing so, the resultant PIP works as a plug-and-play module to enhance existing restoration models for universal image restoration. Extensive experimental results demonstrate the superior performance of PIP on multiple restoration tasks, including image denoising, deraining, dehazing, deblurring, and low-light enhancement. Remarkably, PIP is interpretable, flexible, efficient, and easy-to-use, showing promising potential for real-world applications. The code is available at https://github.com/longzilicart/pip_universal.
In this paper, we introduce a new perspective for improving image restoration by removing degradation in the textual representations of a given degraded image. Intuitively, restoration is much easier on text modality than image one. For example, it can be easily conducted by removing degradation-related words while keeping the content-aware words. Hence, we combine the advantages of images in detail description and ones of text in degradation removal to perform restoration. To address the cross-modal assistance, we propose to map the degraded images into textual representations for removing the degradations, and then convert the restored textual representations into a guidance image for assisting image restoration. In particular, We ingeniously embed an image-to-text mapper and text restoration module into CLIP-equipped text-to-image models to generate the guidance. Then, we adopt a simple coarse-to-fine approach to dynamically inject multi-scale information from guidance to image restoration networks. Extensive experiments are conducted on various image restoration tasks, including deblurring, dehazing, deraining, and denoising, and all-in-one image restoration. The results showcase that our method outperforms state-of-the-art ones across all these tasks. The codes and models are available at \url{https://github.com/mrluin/TextualDegRemoval}.
Among diverse contemporary colour prints, silver dye bleach prints and chromogenic prints are difficult to differentiate. They share similar visual characteristics and can use identical supports and surface finishes. However, their image-forming dyes differ, resulting in disparate conservation and restoration needs. This study aimed to determine practical measures for unambiguously differentiating between these two print types. Identifying characteristics—referred to here as ‘identifiers’—were collected from popular conservation sources and a mixed-method questionnaire survey. The accuracy and feasibility of these identifiers were evaluated against known prints sets. Examinations made use of water droplets, various light sources, digital 3D microscopy, and spectrophotometry. Results dichotomised these identifiers into ‘definite’ or ‘indefinite’ with ‘definite identifiers’ being able to discriminate independently. Only five out of 23 entries were termed definite identifiers. Azo dyes—image dyes of silver dye bleach prints—were established as the only constant definite identifiers of this print type. These findings were integrated into a flowchart to guide differentiation with the main recommendation being to deviate from indefinite identifiers to save time and effort. Parts of this work have been submitted in partial fulfilment of the requirements for a Master of Science degree at the University of Amsterdam in 2020.
Las pinturas del santuario de Nuestra Señora de Guaditoca en Guadalcanal constituyen un ejemplo paradigmático de pintura mural andaluza de finales del XVIII. En este artículo se abordan el análisis del estado de conservación de las pinturas y un diagnóstico pormenorizado, así como el avance de los criterios y metodología para su intervención. Un trabajo que plantea la lectura del programa iconográfico, hasta el momento inédito, desde su reconocimiento arquitectónico e histórico como bien cultural hasta su caracterización y tutela patrimonial. Su necesaria puesta en valor y consiguiente protección se fundamentan en una urgente intervención de carácter científico-técnico.
Conservation and restoration of prints, Architectural drawing and design
Adaptive image restoration models can restore images with different degradation levels at inference time without the need to retrain the model. We present an approach that is highly accurate and allows a significant reduction in the number of parameters. In contrast to existing methods, our approach can restore images using a single fixed-size model, regardless of the number of degradation levels. On popular datasets, our approach yields state-of-the-art results in terms of size and accuracy for a variety of image restoration tasks, including denoising, deJPEG, and super-resolution.
Xiaoming Li, Shiguang Zhang, Shangchen Zhou
et al.
To improve the performance of blind face restoration, recent works mainly treat the two aspects, i.e., generic and specific restoration, separately. In particular, generic restoration attempts to restore the results through general facial structure prior, while on the one hand, cannot generalize to real-world degraded observations due to the limited capability of direct CNNs' mappings in learning blind restoration, and on the other hand, fails to exploit the identity-specific details. On the contrary, specific restoration aims to incorporate the identity features from the reference of the same identity, in which the requirement of proper reference severely limits the application scenarios. Generally, it is a challenging and intractable task to improve the photo-realistic performance of blind restoration and adaptively handle the generic and specific restoration scenarios with a single unified model. Instead of implicitly learning the mapping from a low-quality image to its high-quality counterpart, this paper suggests a DMDNet by explicitly memorizing the generic and specific features through dual dictionaries. First, the generic dictionary learns the general facial priors from high-quality images of any identity, while the specific dictionary stores the identity-belonging features for each person individually. Second, to handle the degraded input with or without specific reference, dictionary transform module is suggested to read the relevant details from the dual dictionaries which are subsequently fused into the input features. Finally, multi-scale dictionaries are leveraged to benefit the coarse-to-fine restoration. Moreover, a new high-quality dataset, termed CelebRef-HQ, is constructed to promote the exploration of specific face restoration in the high-resolution space.
Three-dimensional restoration of complex structural models has become a recognized validation method. Bringing a sedimentary structural model back in time to various deposition stages may also help understand the geological history of a study area and follow the evolution of potential hydrocarbon source rocks, reservoirs and closures. Most current restoration methods rely on finite-element codes which require a mesh that conforms to both horizons and faults, a difficult object to generate in complex structural settings. Some innovative approaches use implicit horizon representations to circumvent meshing requirements. In all cases, finite-element restoration codes depend on elasticity theory which relies on mechanical parameters to characterize rock behavior during the physical unfolding process. In this paper, we present a geometric restoration method based on the mathematical theory provided by the GeoChron framework. No assumption is made on the extent of deformation, nor on the nature of terrains being restored. Equations derived from the theory developed for the GeoChron model ensure model consistency at each restored stage. As the only essential input is a GeoChron model, this restoration technique does not require any specialist knowledge and can be included in any existing structural model-building workflow as a standard validation tool. A model can quickly be restored to any desired stage without providing input mechanical parameters for each layer nor defining boundary conditions, enabling geologists to iterate on the structural model and refine their interpretations until they are satisfied with both input and restored models.
Remote sensing provides valuable information about objects or areas from a distance in either active (e.g., RADAR and LiDAR) or passive (e.g., multispectral and hyperspectral) modes. The quality of data acquired by remotely sensed imaging sensors (both active and passive) is often degraded by a variety of noise types and artifacts. Image restoration, which is a vibrant field of research in the remote sensing community, is the task of recovering the true unknown image from the degraded observed image. Each imaging sensor induces unique noise types and artifacts into the observed image. This fact has led to the expansion of restoration techniques in different paths according to each sensor type. This review paper brings together the advances of image restoration techniques with particular focuses on synthetic aperture radar and hyperspectral images as the most active sub-fields of image restoration in the remote sensing community. We, therefore, provide a comprehensive, discipline-specific starting point for researchers at different levels (i.e., students, researchers, and senior researchers) willing to investigate the vibrant topic of data restoration by supplying sufficient detail and references. Additionally, this review paper accompanies a toolbox to provide a platform to encourage interested students and researchers in the field to further explore the restoration techniques and fast-forward the community. The toolboxes are provided in https://github.com/ImageRestorationToolbox.
We present a novel approach to image restoration that leverages ideas from localized structured prediction and non-linear multi-task learning. We optimize a penalized energy function regularized by a sum of terms measuring the distance between patches to be restored and clean patches from an external database gathered beforehand. The resulting estimator comes with strong statistical guarantees leveraging local dependency properties of overlapping patches. We derive the corresponding algorithms for energies based on the mean-squared and Euclidean norm errors. Finally, we demonstrate the practical effectiveness of our model on different image restoration problems using standard benchmarks.