Publications
Peer-reviewed papers, preprints and datasets. Citation counts live on Google Scholar.
-
When Restoration Becomes the Reference: Reusing Full-Reference IQA in Blind Settings
conference IEEE International Conference on Image Processing (ICIP)
Abstract
Full-reference image quality metrics are consistently stronger than blind ones, but they are unusable in practice because the pristine reference is rarely available. This work closes that gap by synthesising a reference instead of requiring one: a restoration model reconstructs a pseudo-reference from the distorted input, which a full-reference metric can then score.
The resulting pipeline lets well-established full-reference measures operate in genuinely blind settings, and we analyse when the restored reference is faithful enough for the borrowed metric to stay reliable.
-
ARIQA-3DS: A Stereoscopic Image Quality Assessment Dataset for Realistic Augmented Reality
dataset arXiv preprint · 5th Workshop on Computer Vision for Multimedia Spatial Intelligence
Abstract
As AR technologies advance towards immersive consumer adoption, the need for rigorous quality-of-experience assessment becomes critical. Existing datasets often lack ecological validity, relying on monocular viewing or simplified backgrounds that fail to capture visual confusion — the complex perceptual interplay between real and virtual layers.
ARIQA-3DS is the first large stereoscopic AR image quality assessment dataset. Its 1,200 AR viewports fuse high-resolution stereoscopic omnidirectional captures of real-world scenes with diverse augmented foregrounds under controlled transparency and degradation conditions. A subjective study with 36 participants using a video see-through head-mounted display collected both quality ratings and simulator-sickness indicators.
Our analysis shows that perceived quality is primarily driven by foreground degradations and modulated by transparency levels, while oculomotor and disorientation symptoms increase progressively but manageably during viewing. The dataset is released publicly as a benchmark for next-generation AR quality models.
-
Enhancing Content Representation for AR Image Quality Assessment Using Knowledge Distillation
journal IEEE Transactions on Circuits and Systems for Video Technology
Abstract
Augmented reality enriches our perception of reality by overlaying digital content onto physical environments, but visual confusion and classical distortions can leave users uncomfortable. Measuring AR quality of experience is therefore essential — yet data scarcity and the distinctive characteristics of AR make effective metrics hard to build.
This paper presents a deep-learning objective metric designed specifically for AR image quality. The approach has four steps: (1) fine-tuning a self-supervised pre-trained vision transformer to extract prominent features from reference images and distilling that knowledge to improve representations of distorted images, (2) quantifying distortions by computing shift representations, (3) employing cross-attention-based decoders to capture perceptual quality features, and (4) integrating regularisation and label smoothing to address overfitting.
Extensive experiments on the ARIQA dataset show that all model variants — TransformAR, TransformAR-KD and TransformAR-KD+ — outperform existing state-of-the-art methods.
-
TransformAR: A Light-Weight Transformer-Based Metric for Augmented Reality Quality Assessment
journal Signal Processing: Image Communication, vol. 140, art. 117437
Abstract
TransformAR is a lightweight transformer-based metric for perceptual quality assessment of augmented reality content. The journal extension of our MMSP 2024 work, it targets the practical constraint that AR quality has to be estimated on-device, under tight compute budgets, while still tracking human judgement closely.
The architecture keeps the transformer’s ability to relate foreground and background regions — the source of visual confusion in AR — but trims capacity aggressively, and is validated against subjective AR quality datasets.
-
ARaBIQA: A Novel Blind Image Quality Assessment Model for Augmented Reality
conference IEEE International Conference on Image Processing (ICIP), pp. 379–384
Abstract
ARaBIQA is a blind (no-reference) image quality assessment model built specifically for augmented reality content, where the visible image is a composite of a captured real scene and a rendered virtual overlay.
Rather than treating the viewport as an ordinary photograph, the model is designed around the distortions that arise in immersive pipelines — transparency, blending artefacts and the perceptual interference between real and virtual layers — and predicts quality without access to any reference.
-
Lightweight Image Quality Prediction Guided by Perceptual Ranking Feedback
conference IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Abstract
Quality models are usually trained to regress absolute scores, even though human observers are far more consistent when ranking two images than when assigning a number to one. This paper exploits that asymmetry.
We introduce a ranking-guided training strategy for lightweight quality prediction: perceptual ordering feedback is folded into the objective so that the model is rewarded for getting pairwise comparisons right, not just for minimising score error. The result is better calibration and stronger pairwise consistency at a small model size.
-
Towards Light-Weight Transformer-Based Quality Assessment Metric for Augmented Reality
conference IEEE 26th International Workshop on Multimedia Signal Processing (MMSP)
Abstract
A lightweight transformer-based approach to perceptual quality assessment in augmented reality, designed to cut computational cost while preserving prediction performance on subjective AR quality datasets.
This is the conference paper that TransformAR extends: it establishes that a compact transformer, trained with the right supervision, is sufficient to model the interaction between the augmented foreground and the captured background.
-
Shifting Focus: From Global Semantics to Local Prominent Features in Swin-Transformer for Knee Osteoarthritis Severity Assessment
conference 32nd European Signal Processing Conference (EUSIPCO), pp. 1686–1690
Abstract
Imaging diagnostics frequently bottleneck on manual inspection, which introduces delays and inconsistency. Deep learning offers a path to automation, but foundation models in computer vision tend to emphasise global context at the expense of the local structures clinicians actually rely on.
We keep the Swin Transformer’s capacity to capture long-range spatial dependencies while refining the representation of clinically salient regions. A localisation-aware refinement module aligns hierarchical transformer tokens with the final classifier distribution, so local cues stay discriminative across every scale. Validation on two public benchmarks shows improved robustness and precision for Kellgren–Lawrence severity prediction.
-
Do Digital Images Tell the Truth?
chapter Digital Image Security: Techniques and Applications, CRC Press, pp. 247–265
Abstract
Since the advent of digital cameras, editing tools have made content manipulation straightforward. Copy-move forgeries — where a region is duplicated and pasted within the same frame — are particularly hard for human observers to detect, yet they undermine the credibility of visual evidence.
This chapter surveys the state of the art in copy-move detection across feature representations, learning strategies and public datasets. We catalogue low-level descriptors, machine-learning pipelines and deep architectures, and discuss how each responds to common post-processing such as translation, rotation and scaling. We close on the open challenges in building explainable, trustworthy forensic systems.
-
Automatic Diagnosis of Knee Osteoarthritis Severity Using Swin Transformer
conference 20th International Conference on Content-Based Multimedia Indexing (CBMI)
Abstract
Knee osteoarthritis is a widespread condition that causes chronic pain and loss of mobility. Early detection and grading are critical for effective intervention, yet manual assessment remains labour-intensive and inconsistent between readers.
We propose an automated framework built on a Swin Transformer backbone with a multi-head prediction module, capturing both global context and fine-grained biomarkers in radiographic images. The model trains on publicly available datasets labelled with Kellgren–Lawrence scores and incorporates a domain adaptation strategy to mitigate dataset shift. Experiments show reliable severity predictions that generalise across data sources.
-
Diagnostic automatique de la sévérité de la gonarthrose à l'aide de transformeur Swin
conference Colloque GRETSI sur le traitement du signal et des images (in French)
Abstract
French-language presentation of our Swin Transformer framework for automatic Kellgren–Lawrence grading of knee osteoarthritis from radiographs, given at the GRETSI signal and image processing colloquium.
-
Deep-Based Quality Assessment of Medical Images Through Domain Adaptation
conference IEEE International Conference on Image Processing (ICIP), pp. 3692–3696
Abstract
Predicting the quality of multimedia content matters in many fields, but in some — diagnosis from medical imagery among them — a quality metric can directly affect decisions. We propose an efficient, shallow model that predicts medical image quality without a reference, from a small amount of annotated data.
The model is based on convolutional self-attention: it builds a complex representation from relevant local characteristics and slides over the image to interpolate a global quality score. We apply domain adaptation in both unsupervised and semi-supervised settings. Results on a dataset of images with corresponding subjective scores confirm the efficiency of the method and the value of domain adaptation for generalising perceptual quality prediction across multimedia domains.
-
Évaluation de la qualité des images médicales basée sur un apprentissage par adaptation au domaine
conference Colloque GRETSI sur le traitement du signal et des images (in French)
Abstract
French-language presentation of our domain-adaptation approach to no-reference quality assessment of medical images, given at the GRETSI signal and image processing colloquium.
No publications match that filter.
Up-to-date citation metrics are on Google Scholar, and the complete record is on ORCID.