IEEE Transactions on Computational Imaging最新文献

{"title":"Gaussian Is All You Need: A Unified Framework for Solving Inverse Problems via Diffusion Posterior Sampling","authors":"Nebiyou Yismaw;Ulugbek S. Kamilov;M. Salman Asif","doi":"10.1109/TCI.2025.3594988","DOIUrl":"https://doi.org/10.1109/TCI.2025.3594988","url":null,"abstract":"Diffusion models can generate a variety of high-quality images by modeling complex data distributions. Trained diffusion models can also be very effective image priors for solving inverse problems. Most of the existing diffusion-based methods integrate data consistency steps by approximating the likelihood function within the diffusion reverse sampling process. In this paper, we show that the existing approximations are either insufficient or computationally inefficient. To address these issues, we propose a unified likelihood approximation method that incorporates a covariance correction term to enhance the performance and avoid propagating gradients through the diffusion model. The correction term, when integrated into the reverse diffusion sampling process, achieves better convergence towards the true data posterior for selected distributions and improves performance on real-world natural image datasets. Furthermore, we present an efficient way to factorize and invert the covariance matrix of the likelihood function for several inverse problems. Our comprehensive experiments demonstrate the effectiveness of our method over several existing approaches.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1020-1030"},"PeriodicalIF":4.8,"publicationDate":"2025-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144843113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Novel Bound for Fourier Ring Correlation in Resolution Analysis","authors":"Eduardo X. Miqueles;Yuri R. Tonin;Russell D. Luke","doi":"10.1109/TCI.2025.3593881","DOIUrl":"https://doi.org/10.1109/TCI.2025.3593881","url":null,"abstract":"This work proposes a novel data-driven approach for computing the resolution number of a given image using the well-established Fourier Ring/Shell Correlation (<sc>frc</small>/<sc>fsc</small>) technique. The proposed method eliminates the need for the user to select a threshold criterion in a heuristic way, a requirement in current methodologies. To achieve this, the approach leverages linear algebra—specifically, Niculescu’s result—and concepts from information theory, demonstrating that the resolution number is directly linked to the Fisher information derived from each ring or shell in the Fourier domain. As a result, the methodology is entirely data-driven, requiring no prior information about the image under analysis. The mathematical framework’s consistency is validated through numerical experiments and tests with real data from x-ray coherent microscopy, tomography, cryo-EM and confocal microscopy, showing that the newly computed resolution numbers align with conventional metrics derived from the classical <inline-formula><tex-math>$1/2$</tex-math></inline-formula>-bit and <inline-formula><tex-math>$3sigma$</tex-math></inline-formula> thresholds. Furthermore, we highlight that the local resolution of images can vary significantly from the single resolution value typically provided by <sc>frc</small>/<sc>fsc</small>. This observation suggests that resolution maps may provide a more reliable framework for assessing resolution in microscopy.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1047-1058"},"PeriodicalIF":4.8,"publicationDate":"2025-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144887635","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Single-Frame MIMO Radar Velocity Vector Estimation via Multi-Bounce Scattering","authors":"Nishant Mehrotra;Divyanshu Pandey;Upamanyu Madhow;Yasamin Mostofi;Ashutosh Sabharwal","doi":"10.1109/TCI.2025.3594013","DOIUrl":"https://doi.org/10.1109/TCI.2025.3594013","url":null,"abstract":"Radars are widely adopted for autonomous navigation and vehicular networking due to their robustness to weather conditions as compared to visible light cameras and lidars. However, radars currently struggle with differentiating static vs tangentially moving objects within a <italic>single radar frame</i> since both yield the same Doppler along line-of-sight paths to the radar. Prior solutions deploy multiple radar or visible light camera modules to form a multi-“look” synthetic aperture for estimating the single-frame velocity vectors, to estimate tangential and radial velocity components of moving objects leading to higher system costs. In this paper, we propose to exploit multi-bounce scattering from secondary static objects in the environment, e.g., building pillars, walls, etc., to form an effective multi-“look” synthetic aperture for single-frame velocity vector estimation with a <italic>single</i> multiple-input, multiple-output (MIMO) radar, thus reducing the overall system cost and removing the need for multi-module synchronization. We present a comprehensive theoretical and experiment evaluation of our scheme, demonstrating a <inline-formula><tex-math>$4.5 times$</tex-math></inline-formula> reduction in the error for estimating moving objects’ velocity vectors over comparable single-radar baselines.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1005-1019"},"PeriodicalIF":4.8,"publicationDate":"2025-07-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11103510","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144843112","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"MAFDE-Net: Multipath Attention-Fusion-Based Dual-Encoder Network for Undersampled MRI Segmentation","authors":"Zhenyu Huang;Jizhong Duan;Yunshuang Xie;Yu Liu","doi":"10.1109/TCI.2025.3592319","DOIUrl":"https://doi.org/10.1109/TCI.2025.3592319","url":null,"abstract":"Magnetic Resonance Imaging (MRI) plays a crucial role in medical diagnosis, but previous studies have mainly relied on fully-sampled magnitude images for segmentation. However, prolonged k-space acquisition may cause discomfort and motion artifacts in patients, and undersampling techniques are commonly used to address these issues. Conventional methods often adopt a strategy of first reconstructing and then segmenting, but this approach neglects the influence of reconstruction on the downstream segmentation task. In view of this, integrating undersampled MRI reconstruction with segmentation and improving undersampled segmentation performance via joint training has emerged as a promising strategy. Therefore, we propose a novel network, MAFDE-Net, that integrates reconstruction and segmentation into a unified framework. The network enhances undersampled MRI segmentation performance through a joint learning mechanism. The proposed framework integrates the R2N branch containing four Res2Net modules, the CNN-Transformer (CT) branch, and the Multipath Attention-Fusion (MAF) module synergistically combining features from both branches. In addition, we include an Inverted Residual (IR) module in the decoder stage to effectively integrate features extracted during the encoding stage. The Dynamic Upsampling (DU) module is introduced to enhance the final upsampling quality. Simulation experiments show that the undersampled segmentation performance of MAFDE-Net on three datasets significantly outperforms the joint model (RecSeg) and seven baseline models (considering reconstruction and segmentation as serial tasks). Additionally, the joint learning mechanism adopted by MAFDE-Net is not limited to undersampling scenarios; it also outperforms single-task models in fully-sampled MRI segmentation tasks, expanding its application scenarios and potential impact.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1099-1114"},"PeriodicalIF":4.8,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144909342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Learned Discrepancy Reconstruction and Benchmark Dataset for Magnetic Particle Imaging","authors":"Meira Iske;Hannes Albers;Tobias Knopp;Tobias Kluth","doi":"10.1109/TCI.2025.3592335","DOIUrl":"https://doi.org/10.1109/TCI.2025.3592335","url":null,"abstract":"Magnetic Particle Imaging (MPI) is an emerging imaging modality based on the magnetic response of superparamagnetic iron oxide nanoparticles to achieve high-resolution and real-time imaging without harmful radiation. One key challenge in the MPI image reconstruction task arises from its underlying noise model, which does not fulfill the implicit Gaussian assumptions that are made when applying traditional reconstruction approaches. To address this challenge, we introduce the Learned Discrepancy Approach, a novel learning-based reconstruction method for inverse problems that includes a learned discrepancy function. It enhances traditional techniques by incorporating an invertible neural network to explicitly model problem-specific noise distributions. This approach does not rely on implicit Gaussian noise assumptions, making it especially suited to handle the sophisticated noise model in MPI and also applicable to other inverse problems. To further advance MPI reconstruction techniques, we introduce the MPI-MNIST dataset — a large collection of simulated MPI measurements derived from the MNIST dataset of handwritten digits. The dataset includes noise-perturbed measurements generated from state-of-the-art model-based system matrices and measurements of a preclinical MPI scanner device. This provides a realistic and flexible environment for algorithm testing. Validated against the MPI-MNIST dataset, our method demonstrates significant improvements in reconstruction quality in terms of structural similarity, achieving up to 7.9% higher SSIM as well as 2.2 dB higher PSNR compared to classical reconstruction techniques across varying noise levels, underscoring its robustness in high-noise scenarios.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1059-1073"},"PeriodicalIF":4.8,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11095657","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144904829","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Compressive Radio-Interferometric Sensing With Random Beamforming as Rank-One Signal Covariance Projections","authors":"Olivier Leblanc;Yves Wiaux;Laurent Jacques","doi":"10.1109/TCI.2025.3587449","DOIUrl":"https://doi.org/10.1109/TCI.2025.3587449","url":null,"abstract":"Radio-interferometry (RI) observes the sky at unprecedented angular resolutions, enabling the study of several far-away galactic objects such as galaxies and black holes. In RI, an array of antennas probes cosmic signals coming from the observed region of the sky. The covariance matrix of the vector gathering all these antenna measurements offers, by leveraging the Van Cittert-Zernike theorem, an incomplete and noisy Fourier sensing of the image of interest. The number of noisy Fourier measurements—or <italic>visibilities</i>—scales as <inline-formula><tex-math>$mathcal O(Q^{2}B)$</tex-math></inline-formula> for <inline-formula><tex-math>$Q$</tex-math></inline-formula> antennas and <inline-formula><tex-math>$B$</tex-math></inline-formula> short-time integration (STI) intervals. We address the challenges posed by this vast volume of data, which is anticipated to increase significantly with the advent of large antenna arrays, by proposing a compressive sensing technique applied directly at the level of the antenna measurements. First, this paper shows that <italic>beamforming</i>—a common technique of dephasing antenna signals—usually used to focus some region of the sky, is equivalent to sensing a rank-one projection (ROP) of the signal covariance matrix. We build upon our recent work (Leblanc et al., 2024) to propose a compressive sensing scheme relying on random beamforming, trading the <inline-formula><tex-math>$Q^{2}$</tex-math></inline-formula>-dependence of the data size for a smaller number <inline-formula><tex-math>$P$</tex-math></inline-formula> of ROPs. We provide image recovery guarantees for sparse image reconstruction. Secondly, the data size is made independent of <inline-formula><tex-math>$B$</tex-math></inline-formula> by applying <inline-formula><tex-math>$M$</tex-math></inline-formula> random modulations of the ROP vectors obtained for the STI. The resulting sample complexities, theoretically derived in a simpler case without modulations and numerically obtained in phase transition diagrams, are shown to scale as <inline-formula><tex-math>$mathcal O(K)$</tex-math></inline-formula> where <inline-formula><tex-math>$K$</tex-math></inline-formula> is the image sparsity. This illustrates the potential of the approach.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1229-1242"},"PeriodicalIF":4.8,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145090146","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Laser Protection via Jointly Learned Defocus and Image Reconstruction","authors":"Johannes Meyer;Michael Henrichsen;Christian Eisele;Bastian Schwarz;Jürgen Limbach;Gunnar Ritt;Stefanie Dengler;Lukas Dippon;Christian Kludt","doi":"10.1109/TCI.2025.3587469","DOIUrl":"https://doi.org/10.1109/TCI.2025.3587469","url":null,"abstract":"We propose a method to harden sensors against laser radiation by defocusing the employed optics on purpose, and to reconstruct the sought focused images of the scene via image reconstruction. The introduced defocus widens the laser spot incident on the sensor and greatly reduces its damage potential. We employ a coded aperture and optimize its pattern jointly with the free parameters of the image reconstruction pipeline. For the image reconstruction, we combine a state-of-the-art alternating direction method of multipliers (ADMM)-based physically informed deconvolution stage with a U-Net-like neural network to remove remaining reconstruction artifacts. To evaluate the performance of our proposed approach, we conducted reconstruction experiments on simulated data, including ablation experiments and on real data and performed sensor destruction tests with and without sensor protection. Destructive experiments with increasing laser power suggest that our approach has the potential to increase the tolerable radiation threshold by about three orders of magnitudes.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"967-979"},"PeriodicalIF":4.2,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11074739","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144680867","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Robust Preprocessing of Impulsive Motion Artifacts Using Low-Rank Matrix Recovery for Electrical Impedance Tomography","authors":"Xiao-Peng Li;Zhang-Lei Shi;Meng Dai;Hing Cheung So;Inéz Frerichs;Zhanqi Zhao;Lin Yang","doi":"10.1109/TCI.2025.3587458","DOIUrl":"https://doi.org/10.1109/TCI.2025.3587458","url":null,"abstract":"Electrical impedance tomography (EIT) is a valuable bedside tool in critical care medicine and pneumology. However, artifacts associated with body and electrode movements, especially impulsive motion artifacts, hinder its routine use in clinical scenarios. Most of the existing algorithms for EIT data preprocessing or imaging cannot effectively address this issue. In this paper, we propose a novel method, namely, robust preprocessing for EIT (RP4EIT), to preprocess EIT boundary voltages using the concept of low-rank matrix recovery. It aims to resist impulsive motion artifacts and further to enhance the imaging quality. To attain good performance on both the normal measurements and contaminated data, we design a two-stage denoising algorithm using robust statistical analysis and low-rank recovery. Specifically, EIT boundary voltages are first formulated as a matrix, where the rows and columns correspond to the channels and frames, respectively. Then, the entries corrupted by impulsive noise of the matrix are identified and considered as missing elements. Subsequently, RP4EIT exploits the low-rank property to restore the missing components. In doing so, the impulsive motion artifacts are eliminated from EIT measurements. Furthermore, the convergence guarantee of RP4EIT is established. Experimental results on phantom and patient data demonstrate that RP4EIT is able to remove the impulsive motion artifacts from boundary voltages and the recovered data yield high-quality EIT images.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"942-954"},"PeriodicalIF":4.2,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144671210","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hybrid Spatial and Frequency Network for Light Field Image Restoration","authors":"Vinh Van Duong;Thuc Nguyen Huu;Jonghoon Yim;Byeungwoo Jeon","doi":"10.1109/TCI.2025.3587603","DOIUrl":"https://doi.org/10.1109/TCI.2025.3587603","url":null,"abstract":"This paper proposes a novel hybrid light field (LF) restoration method based on a deep convolutional neural network (CNN) designed to capture the characteristics of LF images in both pixel and frequency domains. Restoring high-quality LF images from degraded versions is a complex task due to the high dimensionality of LF data. To address this, we leverage the geometric priors of LF images to design efficient restoration network components capable of effectively handling the 4D LF structure across both pixel and frequency domains. In the frequency restoration stage, where image artifacts often exhibit distinct frequency characteristics, we propose a 4D-DCT separated transform using 2D-DCT in spatial and angular pixel correlations. By decomposing transformed LF data into various frequency components, our frequency restoration network progressively recovers detailed information from each subband frequency component, enhancing performance in complex scenes and noisy images. For pixel restoration, we introduce the geometry-aware attention (GAM) mechanisms into spatial, angular, and epipolar dimensions of the 4D LF structure, helping to capture better global information in each LF embedding feature. Extensive experiments across diverse LF restoration tasks, including LF denoising, LF spatial super-resolution, and LF low-light enhancement, validate the effectiveness of our method compared to state-of-the-art approaches in both objective and subjective quality assessments.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"1031-1046"},"PeriodicalIF":4.8,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144853399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Test-Time Adaptation Improves Inverse Problem Solving With Patch-Based Diffusion Models","authors":"Jason Hu;Bowen Song;Jeffrey A. Fessler;Liyue Shen","doi":"10.1109/TCI.2025.3587407","DOIUrl":"https://doi.org/10.1109/TCI.2025.3587407","url":null,"abstract":"Diffusion models have achieved excellent success in solving inverse problems due to their ability to learn strong image priors, but existing approaches require a large training dataset of images that should come from the same distribution as the test dataset. In practice, the size of the available training dataset can range from nonexistent to very large. In some cases, conventional diffusion model training from limited data can lead to poor reconstruction results due to poorly learned priors. One potential improvement is to start with a diffusion model trained from available training data having a possibly mismatched distribution, and then refine the network at reconstruction time to account for the distribution mismatch. In this work, we investigate the effect of this network refining process on diffusion models trained from varying degrees of out-of-distribution data. Specifically, we use a self-supervised loss to adapt the learned diffusion network to the testing data while helping the network output maintain consistency with the measurements. We show that, both theoretically and experimentally, test-time adaptation of a patch-based diffusion prior leads to higher quality reconstructions than test-time refinement of traditional whole-image diffusion models. Extensive experiments show that across a wide range of inverse problems, test-time adaptation significantly improves image reconstruction quality when there are significant domain shifts between training and testing distributions. Interestingly, even for the in-distribution case, test-time adaptation also significantly improves reconstruction quality.","PeriodicalId":56022,"journal":{"name":"IEEE Transactions on Computational Imaging","volume":"11 ","pages":"980-991"},"PeriodicalIF":4.2,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144671172","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}