IQA Vision Transformed: A Survey of Transformer Architectures in Perceptual Image Quality Assessment
In an era dominated by visual content, perceptual image quality assessment (IQA) is crucial for enhancing user experiences and driving technological advancements across various domains. This survey paper reviews the integration of Vision Transformers (ViTs) into both no-reference (NR) and full-refer...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10767243/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In an era dominated by visual content, perceptual image quality assessment (IQA) is crucial for enhancing user experiences and driving technological advancements across various domains. This survey paper reviews the integration of Vision Transformers (ViTs) into both no-reference (NR) and full-reference (FR) IQA methods, highlighting their promise as alternatives to traditional techniques. ViTs leverage attention mechanisms to focus selectively on relevant image patches, showing promise in aligning more closely with human perceptual errors. We identify key limitations of conventional IQA methods and track the evolution from early learning-based approaches to contemporary deep learning models, with a specific focus on ViTs. We discuss the performance of Transformer-based models in capturing image distortions and their strong correlation with subjective IQA metrics. We also discuss potential advancements, including the development of hybrid architectures combining diverse deep learning approaches, adaptive IQA mechanisms through meta-learning, and scalable solutions inspired by emerging computational paradigms. These advancements promise to enhance perceptual quality assessment, with substantial implications for industries such as medical imaging, multimedia applications, and beyond. This study aims to set the groundwork for future research in transformer-based methodologies, offering new insights into the transformative impact of these models on IQA. |
|---|---|
| ISSN: | 2169-3536 |