CS Paper Radar

Daily listing

2026-11-19 · eess.IV

14 papers first submitted this day (UTC) in eess.IV — Image and Video Processing. arXiv announces a day's submissions the next evening, so the newest day fills up over ~24 h.

Score = weighted mean ofCitations 20%Upvotes 10%Code stars 10%Venue 15%Code/data 10%Novelty 15%Momentum 15%Substance 5%how it works

AI & Machine Learning 1

MambaLIFT: Reference-Based Mechanism-Aware Attribution for VMamba

cs.AI cs.CVeess.IV ·2026-11-19 ·Jintu Kakati, Niladri B. Puhan
hot direction
Abstract

Attributing predictions of visual state-space models remains challenging because input-dependent selective dynamics couple feature selection, recurrent state propagation, and shared computational pathways. We introduce MambaLIFT, a reference-based, multiplier-first attribution framework for VMamba that propagates contributions through its computational graph using interaction-aware rules for selective-scan injection, recurrent memory, LayerNorm, and shared pathways. MambaLIFT decomposes the reference-relative class-logit difference, \(f_c(x)-f_c(x^0)\), by spatial position, VSS block, scan direction, and computational mechanism. Experiments on ImageNet-1K evaluate VMamba-T and VMamba-B, with hyperparameters selected on a calibration set and faithfulness assessed on a disjoint held-out evaluation set. MambaLIFT achieves numerical conservation to floating-point precision across audited attribution interfaces in both configurations. Under mean-colour pixel flipping, it achieves significantly higher mean faithfulness scores than MambaLRP on both configurations and the highest scores among the evaluated methods, including Integrated Gradients with matched references. Under blurred replacement, MambaLIFT is non-inferior to MambaLRP on VMamba-T and equivalent on VMamba-B, indicating that its advantage depends on the replacement protocol. Ablations on VMamba-T show that interaction allocation, temporal propagation, and pathway composition affect attribution maps. Mechanism-level analyses further show that exact relevance conservation does not imply additive effects under finite interventions: the mechanism decomposition accounts for the reference-relative logit difference but does not estimate intervention effects or establish causal validity.

0.36

Language, Vision & Media 12

Breaking Noise Shortcuts in Self-Supervised Learning Via Noise-Aligned View Generation

eess.IV cs.AIcs.CV ·2026-11-19 ·Oankar R. Patil, May D. Wang
publishedcode/data
Abstract · 40th Conference on Neural Information Processing Systems (NeurIPS 2026)

Self-supervised learning (SSL) is increasingly applied to scientific imaging domains, such as microscopy, where labels are scarce but noisy data is abundant. These domains exhibit significant signal-dependent noise, which can create a representational shortcut: since both augmented views derive from the same noisy image, the encoder can boost view agreement by also encoding noise patterns rather than semantic content. To enable noise-robust representation learning without knowledge of underlying noise models, we introduce Self-Aligned Noise Augmentation (SEANA), a drop-in module that generates noise-aware SSL views. SEANA learns an invertible variance-stabilizing transform (VST) from noisy data, then estimates the clean signal and resamples independent noise in learned VST space, requiring no clean targets, no dataset-specific denoiser training or use at inference, and no changes to the SSL objective or encoder. On CIFAR-10 and ImageNet-100 under synthetic Gaussian, Poisson-Gaussian, and multiplicative noise, SEANA improves clean-test linear-probe accuracy across contrastive/non-contrastive SSL methods, indicating higher-quality representations. On real fluorescence imaging, SEANA improves Jurkat cell-cycle stage classification by up to 28.6pp over standard SSL and 27.5pp over denoiser-preprocessed baselines. In all settings, SEANA outperforms the denoiser-preprocessed SSL pipelines, demonstrating that learned VST-space noise-aligned resampling yields more robust representations than denoising alone. Code is available at: https://github.com/opatil31/SEANA

0.53

LBW-NPS: A Local Band-Weighted Noise Power Spectrum Loss for Texture-Preserving Low-Dose CT Denoising

eess.IV ·2026-11-19 ·Fatma Elif Yalçındağ, Mehmet Ozan Ünal, Metin Ertaş, İsa Yıldırım
code/data
Abstract · 5 pages, 3 figures, 2 tables. Submitted to IEEE ISBI 2027. Code: https://github.com/itu-biai/lbw-nps

Low-dose CT denoising should suppress noise while retaining the local texture of the reference image. Matching a pooled spectrum can overlook regional discrepancies, whereas pixel losses alone can favor overly smooth outputs. We propose a local band-weighted noise power spectrum loss (LBW-NPS) that compares spectral energies within overlapping Hann-tapered patches, without tissue-specific loss conditioning. Each valid patch contributes a weighted squared log-energy difference across three frequency bands, alongside an $L_1$ reconstruction term. We compare LBW-NPS against $L_1$-only training and five auxiliary-loss baselines using the same U-Net and three seeds on DeepLesion slices with simulated low-dose acquisitions, including patch-scale ablations and lung, soft-tissue, and bone evaluations. On 220 test slices from 31 patients, LBW-NPS reduces a local spectral gap related to the loss, on $64\times64$-pixel evaluation windows, by 31.9% relative to pooled band-weighted NPS, without a statistically significant PSNR change. This improvement is supported by lower errors relative to the reference in average NPS frequency and gray-level co-occurrence matrix (GLCM) measures not used in the loss, evaluated in homogeneous soft tissue. Local band-power matching therefore improves regional texture agreement without changing the denoising architecture.

0.52

Physics-Informed Deep Learning for Shear Wave Speed Estimation in MR Elastography

eess.IV cs.LGphysics.med-ph ·2026-11-19 ·Stefan Martin, Jakob Schattenfroh, Patrick Schuenke, Felix Frederik Zimmermann, Ingolf Sack +2 more
top venuecode/data
Abstract · 13 pages (incl. cover page), 6 figures, 2 tables. Accepted version of the article published in IEEE Transactions on Biomedical Engineering. …

Objective: Magnetic Resonance Elastography (MRE) is a non-invasive imaging technique for mapping biomechanical properties of in vivo tissue, including shear wave speed (SWS), but involves intrinsically slow data acquisition and an ill-posed wave inversion. Instead of relying on handcrafted image priors, we propose a data-driven approach jointly combining image reconstruction and MRE inversion for robust SWS estimation from undersampled k-space data. Methods: Our physics-informed reconstruction framework comprises two blocks: a model-based neural network (NN)-regularized reconstruction module and a phase-gradient inversion (k-MDEV) calculating SWS from the reconstructed images. Concatenating both blocks yields an end-to-end trainable method to estimate SWS directly from measured k-space data. We evaluated the method on retrospectively highly undersampled brain MRE data and compared it to a total variation (TV) minimization-based approach. We assessed the impact of end-to-end training (qualitative images and SWS maps as targets) versus pre-training (qualitative images as targets) and applied the method also to in vivo data. Results: Our approach significantly reduces NRMSE by 30% compared to TV. End-to-end training improves SWS estimation over separate image reconstruction and SWS calculation. Conclusion: Accurate SWS quantification is possible at acceleration factors up to 19. Our method significantly outperforms TV, highlighting the need for data-driven regularization in this challenging MR problem. Further, our approach successfully generalizes to in vivo data. Significance: We present the first end-to-end trainable MRE reconstruction method for estimating SWS maps directly from k-space. NN-based reconstruction can enable rapid stiffness mapping for dynamic studies, functional imaging, and real-time clinical feedback.

0.49

AIstatLab at BraTS-GoAT 2026 Task 3: Quality-Controlled Consensus Pseudo-Labeling and Dense Expert Adaptation

eess.IV cs.AI ·2026-11-19 ·Heeseung Shin, Changwon Lim
publishedcode/data
Abstract · 12 pages; BraTS 2026 Challenge, MICCAI 2026

BraTS-GoAT Task 3 evaluates brain-tumor segmentation across heterogeneous entities using shared ET, TC, and WT regions. We present a 3D ResidualEncoderUNet trained with quality-controlled consensus pseudo-labels, a training-only nested-region objective, and a four-expert bottleneck adapter. Five-fold ResEnc-M/L teachers gener ated predictions for 1,138 unlabeled cases; fusion, hierarchy enforcement, and confidence-and-agreement screening retained 983 for training with 1,351 annotated cases. The final five-fold ResEnc-M entry achieved mean DSC 0.834, mean NSD 0.510, and Acalc = 0.6720 on the validation server. Component effects were modest and region-dependent, with the clearest gains on ET. Source code is available at https://github.com/HeeSeung-Shin/Brats2026-Goat-HS.

0.42

OffBeat: Multicoset Sampling for 4D Capture Beyond the Camera Frame Rate

cs.CV eess.IV ·2026-11-19 ·Chi-Wei Lee, An-Kai Chen, Cheng-Han Chiang, Yi-Chun Chang, Po-Chih Kuo
code/data
Abstract · Project page: https://arthur031221.github.io/offbeat/

For 4D reconstruction, a synchronized multi-camera rig records all views at the same timestamps and misses the motion between them. Staggered triggers spread the same frames over more timestamps. However, when the cameras along each row fire one after another, as in 4DSloMo's schedule on the N3DV rig, the row acts as one moving camera that confounds depth with motion. We introduce OffBeat, which models a staggered rig as a spatially distributed multicoset sampler: shifted copies of one sampling grid. From the calibration alone, OffBeat derives camera groups that fire together and their firing order. Since one group records each timestamp, a short-lived primitive can fit an error its cameras share, such as a color offset, and flicker in novel views. We show that a primitive's opacity over time weights all groups equally, for every number of groups, only if it lasts at least one camera frame interval. On Spacetime Gaussians, a minimum lifetime approximates this bound, and a closed-form smoothed box of that length satisfies it exactly, with less flicker than a Gaussian in its place at nearly equal PSNR. In simulated staggered capture on N3DV, OffBeat outperforms the full 4DSloMo pipeline with video diffusion on every test frame, by 3.04 dB in mean dynamic-region PSNR. Between the first and last synchronized frame, it outperforms the best synchronized method of each scene, with or without frame interpolation, by 1.98 dB in mean dynamic-region PSNR, while synchronized capture is better at its own frames. On PanopticSports, in this frame range, OffBeat is comparable in PSNR to synchronized capture with interpolation. On every N3DV scene, replacing 4DSloMo's schedule with ours raises the dynamic-region PSNR of Spacetime Gaussians, 4DSloMo without diffusion, and our reconstruction. Multi-camera rigs could follow faster motion at the same frame count by changing only when each camera fires.

0.40

Life-IQA+: Boosting Blind Image Quality Assessment via GCN-enhanced Layer Interaction and Flexible Hierarchical Expert Specialization

eess.IV cs.CV ·2026-11-19 ·Long Tang, Guoquan Zheng, Huiyu Duan, Jie Hao, Yucheng Zhu +3 more
code/data
Abstract

Blind image quality assessment (BIQA) is essential for evaluating and optimizing visual experience. Existing BIQA methods typically aggregate features from multiple backbone stages without accounting for their unequal contributions to quality prediction. Furthermore, uniformly supervising all experts may constrain their flexibility and limit the diversity of the learned quality representations. To address these issues, we present Life-IQA+, a quality representation decoding framework that combines graph convolutional network (GCN) enhanced layer interaction and flexible hierarchical expert specialization. Specifically, the layer interaction module constructs quality queries from GCN-refined deepest-stage features and uses penultimate-stage features as keys and values, thereby enabling cross-layer representation interaction via the cross-attention mechanism. Building on these representations, we further introduce a flexible hierarchical mixture-of-experts (MoE) architecture comprising a densely activated shared expert layer and a sparsely activated routed expert layer. To enhance the representation diversity and ability of experts, we apply auxiliary classification supervision to a subset of shared experts using distortion-type labels for synthetic datasets and discretized quality-level labels for authentic datasets. Experiments on synthetic and authentic distortion benchmarks, together with controlled ablations of expert configurations and supervision strategies, show that Life-IQA+ achieves competitive prediction accuracy with a favorable balance between accuracy and computational cost. The code is available at: https://github.com/TANGLONG2/Life-IQA-PlusPlus.

0.39

From power to spectrum: an off-the-shelf photodiode turned into a spectrometer by its bias alone

eess.IV physics.opticseess.SPeess.SY ·2026-11-19 ·Roi Yozevitch, Shmuel Sternklar, Ziv Glasser
unusual
Abstract · 15 pages, 9 figures, 5 tables

Computational spectrometers replace dispersive optics with detectors whose spectral response can be varied, but every reported single-detector device obtains this response through dedicated fabrication: two-dimensional heterostructures, graded epitaxial junctions, organic stacks, filters or photonic circuits. We demonstrate a generic, fabrication-free alternative in which a standard commercial photodiode becomes a spectrometer through its electrical drive alone, requiring no modification of the device, no optical filter and no modulation of the light. The mechanism is bias-modulated optoelectronic chromatic dispersion (OED): light of different wavelengths is absorbed at different depths, and a sinusoidal bias changes how carriers generated at each depth are collected, so that every spectrum is encoded in the shape of the photocurrent waveform over one bias period. Applied to an unmodified InGaAs PIN photodiode at room temperature, a 10 us waveform decoded by a neural network reconstructs spectra of one to three laser lines across the 80 nm of the telecom C+L band, locating the lines to 0.3-0.4 nm, counting them in about 99% of measurements and identifying their power levels, with line shifts tracked to tens of picometres. Because OED relies on universal semiconductor transport physics, the approach is material-agnostic and applies in principle to any off-the-shelf photodiode, turning detectors already deployed as power monitors into low-cost spectral sensors.

0.30

Deterministic Temporal Embedding for Calcium Imaging Trace Extraction

cs.CV cs.LGeess.IV ·2026-11-19 ·Qi Gao, Julius Beichert, Miriam Ratliff, Karl Rohr
Abstract · 5 pages. Submitted to IEEE ISBI 2027

Reliable extraction of calcium-dependent fluorescence traces in calcium imaging sequences requires identifying pixels whose fluorescence intensity reflects the same underlying cellular activity. We present a deterministic temporal embedding framework that maps pixel-wise fluorescence traces to a low-dimensional representation in which temporally coherent sources form spatially coherent regions. Starting from a fuzzy-neighborhood graph and spectral initialization, we optimize a class-balanced cross-entropy objective using exact sparse attraction and a deterministic particle-mesh decomposition of the dense repulsive interaction with sparse edge compensation, avoiding stochastic negative sampling. The resulting latent representation is spatially decoded into reliable source supports for trace extraction without imposing an explicit cellular shape prior. Experiments on synthetic and in vivo two-photon calcium imaging data demonstrate robust recovery of irregular and spatially touching sources together with efficient deterministic embedding optimization.

0.27

Susceptibility source separation via Closed-form Alternating Newton-type Optimization

eess.IV ·2026-11-19 ·Ignacio Contreras-Zúñiga, Cristian Tejos, Carlos Milovic
Abstract

Quantitative Susceptibility Mapping (QSM) provides a bulk magnetic susceptibility estimate per voxel, so contributions from paramagnetic and diamagnetic sources within the same voxel partially cancel. Susceptibility source separation methods identify these compartments by fitting a three-pool complex exponential model to multi-echo gradient-echo (GRE) data. However, conventional voxel-wise fitting with general-purpose constrained solvers requires several hours per whole-brain volume and often fails in high-susceptibility regions. We present CANON-QSM (Closed-form Alternating Newton-type Optimization for QSM), a solver that reformulates the subproblems of the three-pool model and solves them alternately via exact closed-form updates and Newton-type steps, yielding a substantial per-voxel computational speedup. Both in silico and in vivo evaluations demonstrate not only a massive reduction in reconstruction time, completing a whole-brain volume in under two minutes, but also improved accuracy and robustness against artifacts in high-susceptibility regions. Finally, an echo-time analysis of the three-pool model shows that source separation is driven by second-order nonlinearities in signal decay and phase evolution rather than first-order bulk estimates.

0.25

FirstCross: Convex Single-Photon 3D Reconstruction from Sparse Views

cs.CV eess.IV ·2026-11-19 ·Chi-Wei Lee, Cheng-Han Chiang, Yi-Chun Chang, An-Kai Chen, Po-Chih Kuo
theorycode/data
Abstract · Project page: https://arthur031221.github.io/firstcross/

Single-photon lidar records the arrival time of each returning photon, enabling 3D reconstruction from few views and scarce photons. However, leading neural methods such as Transientangelo require over an hour of non-convex per-scene optimization, and their accuracy degrades as photon flux decreases. We propose FirstCross, a network-free approach that instead solves a short sequence of convex problems on a voxel grid. For a solid object, the Poisson likelihood of a ray with given signal and background rates depends on the scene solely through its first crossing, the point where the ray first enters the object. Decomposing this likelihood yields two tractable terms. The first penalizes surfaces in front of detected photons and admits a tight convex closure, namely volume rendering with a running-maximum transmittance in place of NeRF's product transmittance. The second accounts for photons that no surface explains and admits a convex upper bound once the object is assumed to be thick behind each detected surface point. Compared with Transientangelo under its own real-data protocol, FirstCross achieves lower depth error across every test scene, view count, and flux level. With three views and 100 times fewer photons than at the protocol's lowest flux, our mean object depth error stays under half of Transientangelo's with all photons. Given our detected surface points, the best classical fusion has 1.3 to 1.8 times our three-view error across flux levels. On real room-scale captures with 1% of their photons, TransientNeRF has 2.8 to 7.7 times our error. FirstCross lets few-view single-photon scanners cut exposure to 1% and reconstruct an object in about a minute on one GPU.

0.23

AnyVIO: Feed-Forward Streaming Visual-Inertial Odometry Across Platforms

eess.IV ·2026-11-19 ·Omoruyi Atekha, Marcus Greiff, Ray Zhang, Luca Carlone
Abstract · 11 pages

Geometric foundation models (GFMs) estimate camera mo- tion and scene structure from images without requiring ex- pert parameter tuning, but either lack metric scale or rely on learned priors to infer scale, which is unobservable from monocular images alone. Inertial sensing offers a way to observe metric scale, but existing methods add inertial in- formation through a filter or factor graph, which reintro- duces bespoke parameter tuning and optimization costs that can grow with the trajectory. To bridge this gap, we present AnyVIO, a real-time, feed-forward streaming model for monocular visual-inertial odometry (VIO), which combines a streaming GFM with an inertial foundation model (IFM). Both the GFM and IFM predict the metric motion between consecutive frames, and a lightweight learned fusion head quantifies the reliability of each estimate and blends them in a single forward pass. Across nine datasets recorded from cars, legged and ground robots, and handheld and head- mounted devices, AnyVIO reduces the metric position error of its visual model by 36% while publishing each pose esti- mate 70 ms after an image is received. We further deploy AnyVIO onboard two real vehicles at up to 140 km/h, where it remains accurate in poor lighting and heavy dust.

0.20

Uncertainty-Aware Multimodal Deep Learning for Prediction of Radiation-Induced Xerostomia in Head and Neck Cancer

eess.IV ·2026-11-19 ·Jin Yang, Tian Liu, Jose Penagaricano, Jing Wang, Erin L. Moshier +3 more
Abstract · 24 pages, 4 figures

To develop and evaluate an uncertainty-aware multimodal learning framework for predicting radiation-induced xerostomia in patients receiving head-and-neck cancer radiotherapy. We developed a Bidirectional Attention-enhanced Diffusion Transformer (BA-DiT) that combines self-supervised multimodal representation learning with supervised Bayesian fine-tuning. The model integrates pretreatment CT images and three-dimensional dose distributions, focusing on bilateral parotid glands while incorporating surrounding anatomical and dosimetric information. During pretraining, bidirectional cross-attention integrates anatomical and dosimetric features, while a latent Diffusion Transformer learns joint representations through diffusion-based learning, feature reconstruction, and contrastive alignment. Supervised fine-tuning employs a Bayesian classifier to predict individualized xerostomia risk and quantify predictive uncertainty, with a learnable label-transition model accounting for potential toxicity label uncertainty. BA-DiT was developed using four independent cohorts (Moffitt, Emory, MSHS, and TCIA) and evaluated against baseline and ablation models. Evaluation included predictive performance, calibration, uncertainty quantification, decision curve analysis, and patient-level case studies. In independent MSHS and Emory cohorts, BA-DiT achieved AUCs of 0.83 and 0.80, respectively, outperforming baseline models. Predictive uncertainty peaked near the decision boundary, with aleatoric uncertainty being the predominant component. Uncertainty-based selective prediction generally improved net benefit among retained patients, with greater gains at lower prediction coverage. BA-DiT enables individualized xerostomia risk prediction while providing complementary uncertainty estimates, potentially supporting informed clinical decision-making and personalized toxicity management.

0.14

Security & Privacy 1

Passing NPCR Without Diffusion: Exact Signatures, a Detection Benchmark and an Audit of Published Image Ciphers

cs.CR cs.MMeess.IV ·2026-11-19 ·Mua'ad Alfarajat
publishedbenchmarkunusual
Abstract · Accepted for publication in the Journal of Information Security and Applications. This is the submitted version; the accepted manuscript wil…

NPCR, the share of ciphertext bytes changed by a one-pixel plaintext change, is reported for nearly every image cipher as evidence of diffusion, with 99.61% the target. Several mechanisms produce values near it, and the mean does not say which. We separate three regimes: no propagation, one round of forward and backward chaining linear over addition or XOR, and independent ciphertexts. For the second an exact count applies. After a least-significant-bit change its mean asymptotically matches the independent-ciphertext reference, while its spread without cancellation is about five to twenty times larger. In simulation, the one-sided NPCR test of Wu, Noonan and Agaian flags one-round additive, mixed and S-box chaining in 2-4% of papers; read in both directions, the same five values flag at least 99.9%. Two-sided testing does the detecting; the spread and the other signatures help diagnose the structure. We then audit twenty-one published schemes. The count matches a reimplementation of one published diffusion layer on every trial and accounts for its reported table, the signatures show two schemes whose printed specifications cannot produce their reported values, and find that three of our own report plain-versus-cipher values and fall to two chosen images; ten more change their encryption material with the plaintext or per encryption, so their NPCR mostly reflects that change. A released checker applies the signatures to published tables, showing what a table measures within stated assumptions; it does not certify security.

0.64