arXiv:2601.08341v1 Announce Type: new 
Abstract: Single Image Super-Resolution (SISR) is a fundamental computer vision task that aims to reconstruct a high-resolution (HR) image from a low-resolution (LR) input. Transformer-based methods have achieved remarkable performance by modeling long-range dependencies in degraded images. However, their feature-intensive attention computation incurs high computational cost. To improve efficiency, most existing approaches partition images into fixed groups and restrict attention within each group. Such group-wise attention overlooks the inherent asymmetry in token similarities, thereby failing to enable flexible and token-adaptive attention computation. To address this limitation, we propose the Individualized Exploratory Transformer (IET), which introduces a novel Individualized Exploratory Attention (IEA) mechanism that allows each token to adaptively select its own content-aware and independent attention candidates. This token-adaptive and asymmetric design enables more precise information aggregation while maintaining computational efficiency. Extensive experiments on standard SR benchmarks demonstrate that IET achieves state-of-the-art performance under comparable computational complexity.

تم تقديم المحول الاستكشافي الفردي (IET) كنهج جديد لتحسين دقة الصورة الفردية (SISR)، مما يعزز كفاءة آليات الانتباه في إعادة بناء الصور من خلال السماح لكل رمز باختيار مرشحي الانتباه المناسبين لمحتواه. تعالج هذه التطورات القيود المفروضة على طرق الانتباه التقليدية التي تتجاهل أوجه التشابه بين الرموز.

Se ha introducido el Transformer Exploratorio Individualizado (IET) como un enfoque novedoso para la Super-Resolución de Imágenes Únicas (SISR), mejorando la eficiencia de los mecanismos de atención en la reconstrucción de imágenes al permitir que cada token seleccione sus propios candidatos de atención adaptados al contenido. Este avance aborda las limitaciones de los métodos de atención por grupos tradicionales que pasan por alto las similitudes entre tokens.

Le Transformer Exploratoire Individualisé (IET) a été introduit comme une nouvelle approche pour la Super-Résolution d'Image Unique (SISR), améliorant l'efficacité des mécanismes d'attention dans la reconstruction d'images en permettant à chaque jeton de sélectionner ses propres candidats d'attention adaptés au contenu. Cette avancée répond aux limitations des méthodes d'attention par groupes traditionnelles qui négligent les similarités entre jetons.

The Individualized Exploratory Transformer (IET) has been introduced as a novel approach to Single Image Super-Resolution (SISR), enhancing the efficiency of attention mechanisms in image reconstruction by allowing each token to select its own content-aware attention candidates. This advancement addresses the limitations of traditional group-wise attention methods that overlook token similarities.

From Local Windows to Adaptive Candidates via Individualized Exploratory: Rethinking Attention for Image Super-Resolution

arXiv:2204.13620v5 Announce Type: replace-cross 
Abstract: Single image super-resolution (SISR) has played an important role in the field of image processing. Recent generative adversarial networks (GANs) can achieve excellent results on low-resolution images. However, there are little literatures summarizing different GANs in SISR. In this paper, we conduct a comparative study of GANs from different perspectives. We begin by surveying the development of GANs and popular GAN variants for image-related applications, and then analyze motivations, implementations and differences of GANs based optimization methods and discriminative learning for image super-resolution in terms of supervised, semi-supervised and unsupervised manners, where these GANs are analyzed via integrating different network architectures, prior knowledge, loss functions and multiple tasks. Secondly, we compare the performances of these popular GANs on public datasets via quantitative and qualitative analysis in SISR. Finally, we highlight challenges of GANs and potential research points for SISR.

استعراض حديث حول الشبكات التنافسية التوليدية (GAN) لتحسين الصورة الفردية (SISR) يسلط الضوء على التقدم في معالجة الصور، مع التركيز على تطبيقات GAN المختلفة وأدائها المقارن على مجموعات البيانات العامة. تؤكد الدراسة على نقص الأدبيات الشاملة التي تلخص هذه التطورات، والتي تعتبر حاسمة لتحسين الصور ذات الدقة المنخفضة.

Una reciente encuesta sobre Redes Generativas Antagónicas (GAN) para la Super-Resolución de Imagen Única (SISR) destaca los avances en el procesamiento de imágenes, centrándose en diversas implementaciones de GAN y su rendimiento comparativo en conjuntos de datos públicos. El estudio enfatiza la falta de literatura completa que resuma estos desarrollos, que son cruciales para mejorar imágenes de baja resolución.

Une récente enquête sur les réseaux antagonistes génératifs (GAN) pour la super-résolution d'image unique (SISR) met en lumière les avancées dans le traitement d'image, en se concentrant sur diverses implémentations de GAN et leur performance comparative sur des ensembles de données publiques. L'étude souligne le manque de littérature complète résumant ces développements, qui sont cruciaux pour améliorer les images de basse résolution.

A recent survey on Generative Adversarial Networks (GANs) for Single Image Super-Resolution (SISR) highlights the advancements in image processing, focusing on various GAN implementations and their comparative performance on public datasets. The study emphasizes the lack of comprehensive literature summarizing these developments, which are crucial for enhancing low-resolution images.

Generative Adversarial Networks for Image Super-Resolution: A Survey

One More Thing in AI – Your Shortcut to AI Mastery

From Local Windows to Adaptive Candidates via Individualized Exploratory: Rethinking Attention for Image Super-Resolution

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

Attentive AI

LexiStock AI

ImgUpscaler AI

Novaheadshot

Ready to build your own newsroom?