Image Quality Assessment Techniques and Applications

Image quality assessment (IQA) encompasses a broad array of methods designed to quantify the perceptual fidelity of images and video. At its core lies subjective assessment, wherein human observers rate visual stimuli under controlled conditions. In contrast, objective metrics seek to predict such judgements algorithmically, typically by comparing a distorted image against a pristine reference (full-reference IQA), by modelling known distortions (reduced-reference IQA), or by inferring quality solely from the distorted image (no-reference IQA). Classical measures such as peak signal-to-noise ratio and structural similarity index have yielded valuable insights into compression and transmission artefacts, while more recent approaches leverage natural scene statistics and machine learning to capture complex, context-dependent degradations. Deep convolutional neural networks now underpin many state-of-the-art algorithms, enabling joint learning of local quality features and their relative importance across spatial regions. Applications span digital imaging pipelines, medical diagnostics, satellite and aerial imagery, multimedia streaming and user-generated content platforms. Emerging domains such as virtual and augmented reality, high dynamic range and light field imaging pose new challenges for accurate and efficient quality prediction. The global significance of IQA is reflected in its role in optimising encoding standards, guiding display technologies and ensuring user satisfaction across diverse end-uses. Ongoing research continues to expand the frontiers of perceptually driven evaluation, seeking robust, interpretable and real-time solutions that align closely with human visual perception.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Image Quality Assessment Techniques and Applications publication trend

The graph below shows the total number of articles in image quality assessment techniques and applications across all publications each year (not limited to Nature Index journals).

Technical terms

Full-reference image quality assessment (FR-IQA): An objective approach that compares a distorted image with an undistorted reference to quantify perceptual differences.

No-reference image quality assessment (NR-IQA): A blind method that predicts visual quality solely from the distorted image without access to a reference.

Subjective assessment: Human evaluation of image quality under controlled viewing conditions, providing ground truth for metric validation.

Convolutional neural network (CNN): A class of deep learning models that repeatedly applies convolution and pooling operations to learn hierarchical visual features.

Structural similarity index (SSIM): An objective metric that measures perceived image quality by comparing luminance, contrast and structural information between two images.

References

1. Deep Neural Networks for No-Reference and Full-Reference Image Quality Assessment. IEEE Transactions on Image Processing (2017).
2. No-reference color image quality assessment: from entropy to perceptual quality. EURASIP Journal on Image and Video Processing (2019).
3. RAPIQUE: Rapid and Accurate Video Quality Prediction of User Generated Content. IEEE Open Journal of Signal Processing (2021).
4. Perceptual video quality assessment: a survey. Science China Information Sciences (2024).