Modeling Visual Attention for Enhanced Image and Video Processing Applications
International Journal of Theoretical & Applied Computational Intelligence · 2025
DOI: 10.65278/ijtaci.2025.11 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Theoretical & Applied Computational Intelligence, 2025. Verfasst von Uzair Ishtiaq und 2 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 36 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Visual attention modeling aims to identify image or video regions that are likely to attract human attention and to use that information to support downstream computer-vision tasks. Classical saliency methods rely on low-level cues such as contrast, color, luminance, orientation, rarity, and spatial structure, whereas recent approaches increasingly use deep learning, attention mechanisms, multimodal representations, and temporal modeling. This review revisits the conceptual distinction between bottom-up and top-down attention, summarizes representative computational saliency models, and discusses their use in video summarization, salient-object detection, and perceptual image/video quality assessment. The review also places the classical models used in the experimental comparison Saliency Using Natural Image Statistics (SUN), Spectral Whitening (SW), Edge Distance Saliency (EDS), and Multiscale Contrast Conspicuity (MCC) in the context of current research. The experimental results on the TNO Human Factors Search_2 dataset are retained to provide a direct comparison with the reported human-observer measures, while recent literature is incorporated to explain how contemporary deep and multimodal methods extend the classical framework. In addition, the review surveys state-of-the-art developments up to 2025, including transformer-, diffusion-, and Mamba-based saliency predictors, foundation-model-based salient-object detection, large-language-model-driven video summarization, and large-multimodal-model-based image/video quality assessment, and it outlines open challenges for future research.
Zitieren
APA
Ishtiaq, U., Baig, A. K., & Ishtiaque, Z. (2025). Modeling Visual Attention for Enhanced Image and Video Processing Applications. International Journal of Theoretical & Applied Computational Intelligence. https://doi.org/10.65278/ijtaci.2025.11
Vancouver
Ishtiaq U, Baig AK, Ishtiaque Z. Modeling Visual Attention for Enhanced Image and Video Processing Applications. International Journal of Theoretical & Applied Computational Intelligence. 2025. doi:10.65278/ijtaci.2025.11
BibTeX
@article{Ishtiaq2025,
author = {Ishtiaq, Uzair and Baig, Ajaz Khan and Ishtiaque, Zubair},
title = {Modeling Visual Attention for Enhanced Image and Video Processing Applications},
journal = {International Journal of Theoretical & Applied Computational Intelligence},
year = {2025},
doi = {10.65278/ijtaci.2025.11}
}
EUVIMED