Image and Video Processing Papers of the Week

Integrating two memristors with a logic circuit, we designed a low-power edge detection unit using dual random number hardware computation.

Foundation and zero-/few-shot models were almost absent from the reviewed studies, and most studies measure benchtop accuracy rather than the full inspection-to-actuation pipeline. This review maps where the field has matured and where it has not, offering a structured reference point and a research agenda for non-destructive quality evaluation of these crops.

A gated hierarchical estimator first fits ATG together with radiometric perturbations and selectively introduces global geometry and specimen-warp models only when a reduced chi-squared adequacy test and Bayesian information criterion justify the added degrees of freedom.

This work introduces three lightweight enhancement modules for medical lesion segmentation to improve clinical applicability, particularly under data-scarce and computationally constrained environments.

We present DINOFAS, a DINOv3-based framework that formulates FAS as bonafide-centric representation learning.

The segmentation stage provides anatomically defined regions for radiomic feature extraction, while explainable representations maintain anatomical traceability. A key contribution of this work is its dual evaluation strategy within a unified structural feature space: it addresses both broad transdiagnostic four-class classification (Healthy Control (HC) vs. ADHD vs. BD vs. SZ) and focused binary patient-control contrasts (SZ vs. HC and BD vs. HC).

Future research may increasingly focus on multimodal foundation models capable of generalizable segmentation across diverse clinical scenarios.

We trained a convolutional network and a vision transformer on NIH ChestX-ray14 and evaluated 303,768 decisions per model at the training hospital and at two external sites, in the USA and Vietnam, with radiologist-consensus labels.

While glomerular subtype ratios provide a compact and clinically meaningful summary of glomerular composition, representing pathomic information using whole-slide image (WSI)-level mean values diminishes glomerulus-level heterogeneity and the contribution of less prevalent patterns. In this proof-of-concept study, we investigated whether modeling individual glomeruli using multiple-instance learning (MIL) could provide complementary information to conventional subtype ratios for estimated glomerular filtration rate (eGFR) classification.

However, under tight uplink budgets, existing task-oriented joint source--channel coding (JSCC) and distributed information bottleneck (IB) methods evaluate rate--distortion trade-offs per terminal, tending to undervalue cross-terminal synergistic features and prune them prematurely. To address this issue, we propose a synergy-credited distributed JSCC (\method{}) framework that introduces a differentiable, post-equalization matrix Rényi synergy credit as a surrogate to reward complementary task evidence surviving the wireless channel and backpropagates its gradients to local symbol gates.

Data-driven reconstruction has overcome similar limits in related inverse problems, but its use for ULF-MRI has been hindered by a lack of open datasets, benchmarks, and flexible simulators. To close this gap, we present OpenULF, a modular, open-source simulation toolkit that generates paired training data and supports end-to-end machine-learning (ML) reconstruction.

Experts annotated 25 studies using both methods, and spreadsheet annotations were audited for predefined data-entry errors.

In this study, we present a proof-of-concept framework for DNA-based RGB image storage and reconstruction that integrates marker-aware codebook construction, consensus-based marker-guided decoding from a single sequence, and AI-assisted image restoration.

The presented work considers three approaches for detecting the coordinates of contacts in printed circuit boards (PCBs).

Here we present a new physics-based method for starlight subtraction.

Two students initialized from the same base VLM are finetuned on the same images, using numbers generated by either the biased or control teacher. Across VLMs ranging from 2B to 7B parameters, we find that both animal and facial preferences can transfer through this unrelated task.

The digital texture color card proposed in this study can be applied to virtual clothing displays, which is helpful for designers in visually presenting the combination effects of different textures and colors.