A Survey on IQA: Insights, Analysis, and Future Outlook

🔍 Introduction

Image Quality Assessment (IQA) is a critical component of image-related technologies and plays a pivotal role in the advancement of image processing and computer vision. In recent years, the proliferation of novel training frameworks and machine learning models has given rise to a multitude of IQA methodologies. This survey conducts a comprehensive review of nearly 200 IQA-related publications, synthesizing key developments in the field and systematically categorizing existing approaches based on their underlying models, training frameworks, publication timelines, application scenarios, and academic impact. This structured analysis aims to facilitate a swift introduction for newcomers while providing seasoned researchers with a clearer perspective on the current state of the field. Moreover, we offer a critical evaluation of the advantages and limitations of various IQA methods and present our perspectives on future research directions. To complement the survey, this repository compiles both the IQA techniques discussed in the paper and other approaches that could not be included due to space constraints, thereby serving as a valuable resource to support further advancements in IQA research.

图像质量评估（IQA）在图像相关的技术中起着非常重要的作用，对于图像处理和计算机视觉领域的技术发展有着深远的影响。近年来，随着新型训练框架和机器学习模型的出现，许多IQA方法涌现出来。本综述通过调研接近二百篇IQA相关论文，总结了IQA发展中值得关注的工作，并按照不同方法所用的模型、训练框架、发表时间、使用场景和影响力进行了整理，以方便初学者快速入门、资深研究者更好地了解领域发展现状。随后，我们对诸多IQA方法的优缺点进行了分析，对IQA方法的未来发展提出了自己的见解。本repository旨在列出论文中提到的、以及受篇幅限制未提到的经典IQA方法，配合论文阅读，以促进IQA技术的发展。

🚀English Paper link: A Survey on Image Quality Assessment: Insights, Analysis, and Future Outlook
🚀Chinese Paper link: 在计算机视觉领域中，我们应该如何评估图像质量（万字长文）？
🚀The simplified chinese version: 什么样的图像才是好的图像？近200篇文献总结图像质量检测的最新进展与挑战

Update Records

• 🔥 [Mar.2,2025] The chinese version of our survey has been released on zhihu.
• 🔥 [Feb.12,2025] The first version of our survey has been released on arXiv.

📜 Table of Contents

• 📚 Image Quality Assessment Methods

  • 📗 General Scene Methods

    • 📕 Statistics Methods
      • 📘 HVS-based Methods
      • 📘 Transform Domain-based Methods
      • 📘 Natural Scene Statistics-based Methods
    • 📕 Machine Learning-based Methods
      • 📘 Model-based Methods
        • 📙 Traditional Machine Learning Methods
        • 📙 CNN-based Methods
        • 📙 Transformer-based Methods
      • 📘 Framework-based Methods

  • 📗 Specific Scene Methods

    • 📕 Medical IQA
    • 📕 IQA for Dehazing Algorithms
    • 📕 Portrait Quality Assessment
    • 📕 Specific Distortion
      • 📙 Blue
      • 📙 JPEG compression

👨‍💻 Contributers

• Zhengyi Shi@11xiaoyi11
• Chengqian Ma@step-out
• Weixuan Jin@axax11344

💡 Citation

• If you find this survey useful, please consider citing it.

📧 Contact Us

📚 Image Quality Assessment Methods

Metrics	Citation number as of Feb.1,2025.	Earliest publication time	With/Without reference	Paper name
📖HVS-Based Method
signal-to-noise ratio (SNR)			With
peak SNR (PSNR)			With
mean squared error (MSE)			With
structural similarity (SSIM)	58352	2004.04	With	Image quality assessment: From error visibility to structural similarity
multi-scale SSIM (MS-SSIM)	8043	2004.05	With	Multiscale structural similarity for image quality assessment
universal image quality index (UQI)	7419	2002.03	With	A universal image quality index
feature similarity index (FSIM)	5512	2011.01	With	FSIM: A feature similarity index for image quality assessment
Visual information fidelity (VIF)	4861	2006.02	With	Image information and visual quality
Most Apparent Distortion (MAD)	2250	2010.01	With	Most apparent distortion: Full-reference image quality assessment and the role of strategy
Gradient Magnitude Similarity Deviation (GMSD)	1716	2013.12	With	Gradient magnitude similarity deviation: A highly efficient perceptual image quality index
Visual SNR (VSNR)	1578	2007.08	With	VSNR: A wavelet-based visual signal-to-noise ratio for natural images
Information content Weighted SSIM (IW-SSIM)	1533	2010.11	With	Information content weighting for perceptual image quality assessment
	1319	2002.05	With	Why is image quality assessment so difficult?
Visual Saliency-induced Index(VSI)	1082	2014.08	With	VSI: A visual saliency-induced index for perceptual image quality assessment
GSIM	872	2011.11	With	Image quality assessment based on gradient similarity
complex wavelets-SSIM (CW-SSIM)	455	2005.05	With	Translation insensitive image similarity in complex wavelet domain
HVS-based peak SNR (PSNR-HVS)	433	2006.01	With	A new full-reference quality metrics based on HVS
RR-SSIM	320	2012.08	With	Reduced-reference image quality assessment by structural similarity estimation
PSIM	238	2017.05	With	A fast reliable image quality predictor by fusing micro- and macro-structures
📖Transform Domain-Based Method
BLIINDS-II	1912	2012.03	Without	Blind image quality assessment: A natural scene statistics approach in the DCT domain
	600	2005.03	With	Reduced-reference image quality assessment using a wavelet-domain natural image statistic model
	482	2006.02		An SVD-based grayscale image quality measure for local and global assessment
SFF	180	2013.06	With	Sparse feature fidelity for perceptual image quality assessment
	55	2015.07	With	Image quality assessment: a sparse learning way
QASD	50	2016.06	With	Sparse representation-based image quality index with adaptive sub-dictionaries
	50	2017.11	With	From sparse coding significance to perceptual quality: a new approach for image quality assessment
📖NSS-Based Method
	3526	2006.10		A statistical evaluation of recent full reference image quality assessment algorithms
IFC	1721	2005.11	With	An information fidelity criterion for image quality assessment using natural scene statistics
Tone-Mapped images Quality Index (TMQI)	720	2012.10	With	Objective quality assessment of tone-mapped images
	397	2007.09		Blind image quality assessment through anisotropy
Dynamic Range Independent quality Measure (DRIM)	371	2008.08	With	Dynamic range independent image quality assessment
📖Traditional Machine Learning Method
BRISQUE	5733	2012.08	Without	No-reference image quality assessment in the spatial domain
DIIVINE	2003	2011.01	Without	Blind image quality assessment: From natural scene statistics to perceptual quality
IL-NIQE	1214	2015.04	Without	A feature-enriched completely blind image quality evaluator
	292	2009.03	Without	No-reference image quality assessment using modified extreme learning machine classifier
Multi-Method Fusion (MMF)	216	2012.12	With	Image quality assessment using multi-method fusion
	200	2010.01		Objective image quality assessment based on support vector regression
SVDR	184	2011.09	With	SVD-based quality metric for image and video using machine learning
	176	2015.03	Without	Blind Image Quality Assessment Based on Multichannel Feature Fusion and Label Transfer
ParaBoost	86	2015.12	With	A paraboost method to image quality assessment
📖CNN-Based Method
	12353	2018.01		The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
IQA-CNN	1406	2014.09	Without	Convolutional neural networks for no-reference image quality assessment
	1229	2017.10	FR+NR	Deep neural networks for no-reference and full-reference image quality assessment
DISTS	847	2020.12	With	Image Quality Assessment: Unifying Structure and Texture Similarity
BIQA by a Self-Adaptive Hyper Network	667	2020.06	Without	Blindly Assess Image Quality in the Wild Guided by a Self-Adaptive Hyper Network
MEON	593	2017.11		End-to-end blind image quality assessment using deep neural networks
RankIQA	556	2017.07	Without	RankIQA: Learning from rankings for no-reference image quality assessment
BIECON	487	2016.12	Without	Fully deep blind image quality predictor
	433	2014.08		Blind image quality assessment via deep learning
DeepBIQ	396	2016.02	Without	On the use of deep learning for blind image quality assessment
Deep image Quality Assessment (DeepQA)	304	2017.10	With	Deep learning of human visual sensitivity in image quality assessment framework
DeepSim	190	2017.09	With	DeepSim: deep similarity for image quality assessment
	126	2014.09	Without	Blind image quality assessment using semi-supervised rectifier networks
Re-IQA	84	2023.04	Without	Re-IQA: Unsupervised Learning for Image Quality Assessment in the Wild
Multi-Pooled Inception Features for NR IQA	44	2020.02	Without	Multi-Pooled Inception Features for No-Reference Image Quality Assessment
IQMA Network	27	2021.06	With	IQMA Network: Image Quality Multi-Scale Assessment Network
📖Transformer-Based Method
MUSIQ	601	2021.08	Without	MUSIQ: Multi-Scale Image Quality Transformer
ViT with relative ranking and self-consistency	307	2021.08	Without	No-reference image quality assessment via transformers, relative ranking,and self-consistency
Maniqa	301	2022.04	Without	Maniqa: Muli-dimension attention network for no-reference image quality assessment
TRIQ	230	2020.12	Without	Transformer for Image Quality Assessment
IQT	167	2021.04	With	Perceptual Image Quality Assessment With Transformers
NTIRE 2021 Challenge on Perceptual IQA	114	2021.05		NTIRE 2021 Challenge on Perceptual Image Quality Assessment
NTIRE 2022 Challenge on Perceptual IQA	113	2022.06		NTIRE 2022 Challenge on Perceptual Image Quality Assessment
📖Framework-Based Method
QAC	470	2013.06	Without	Learning without human scores for blind image quality assessment
MetalQA	405	2020.04	Without	MetalQA: Deep Meta-Learning for No-Reference Image Quality Assessment
	315	2018.06		Deep CNN-based blind image quality predictor
UNIQUE	286	2021.03	Without	Uncertainty-Aware Blind Image Quality Assessment in the Laboratory and Wild
Hallucinated-IQA	273	2018.04	Without	Hallucinated-IQA: No-reference image quality assessment via adversarial learning
CONTRIQUE	206	2022.06	Without	Image quality assessment using contrastive learning
DeepFL-IQA	67	2020.01	Without	DeepFL-IQA: Weak Supervision for Deep IQA Feature Learning
CVRDK-IQA	34	2022.02	Without	Content-Variant Reference Image Quality Assessment via Knowledge Distillation
CNN-Based Medical Ultrasound IQA	28	2021.07	Without	CNN-Based Medical Ultrasound Image Quality Assessment
📖Medical IQA
	360	2018.04	Without	Quantitative assessment of structural image quality
📖IQA for Dehazing Algorithm
	222	2019.02	With	Quality evaluation of image dehazing methods using synthetic hazy images
📖Portrait Quality Assessment
	15	2024.04	Without	Deep Portrait Quality Assessment. A NTIRE 2024 Challenge Survey
📖IQA for Low Light Enhancement
NLIEE	55	2021.06	Without	A no-reference evaluation metric for low-light image enhancement
📖Specific Distortion
	1004	2009.04	Without	A no-reference objective image sharpness metric based on the notion of just noticeable blur (JNB)
BIBLE	280	2015.01	Without	No-reference image blur assessment based on discrete orthogonal moments
	127	2013.11		A no-reference metric for evaluating the quality of motion deblurring
	40	2015.02	Without	Full reference image quality metrics for JPEG compressed images

👨‍💻 Contributers

• Zhengyi Shi@11xiaoyi11
• Chengqian Ma@step-out
• Weixuan Jin@axax11344

💡 Citation

If you find this survey useful, please cite our paper:

@misc{ma2025surveyimagequalityassessment,
      title={A Survey on Image Quality Assessment: Insights, Analysis, and Future Outlook}, 
      author={Chengqian Ma and Zhengyi Shi and Zhiqiang Lu and Shenghao Xie and Fei Chao and Yao Sui},
      year={2025},
      eprint={2502.08540},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2502.08540}, 
}

📧 Contact Us

If you have any questions or suggestions, please contact us via:

• Email: chengqianma@yeah.net, zyshi@stu.xmu.edu.cn