Perceptual Visual Quality Assessment: Principles, Methods, and Future Directions

Wei Zhou, Hadi Amirpour, Christian Timmerer, Guangtao Zhai, Patrick Le Callet, Alan C. BovikarXiv 2025 · DOI · arXiv

Introduction

Perceptual Visual Quality Assessment (PVQA) is the interdisciplinary field dedicated to developing computational models that can evaluate images and videos in a manner consistent with human perception. While traditional metrics like Peak Signal-to-Noise Ratio (PSNR) rely on pixel-level mathematical differences, they often fail to capture the subjective experience of a human viewer. As digital media evolves toward immersive virtual reality (VR) and generative AI (GenAI), the need for sophisticated, human-centric quality assessment has never been greater.

Framework of PVQA

The field of PVQA is broadly divided into two methodologies: subjective and objective. Subjective methods involve human observers rating content to establish a "gold standard," typically expressed as a Mean Opinion Score (MOS). Objective methods, conversely, are algorithms designed to predict these human scores automatically.

Overview of Perceptual Visual Quality Assessment (PVQA) components, including subjective and objective methods influenced by multimedia sources and the human visual system.

As shown in the figure above, objective methods are further categorized based on the availability of a reference signal:

  • Full-Reference (FR): Compares a distorted image to a perfect original (e.g., SSIM, VMAF).
  • Reduced-Reference (RR): Uses only partial features of the original.
  • No-Reference (NR): Evaluates quality without any original "ground truth," which is essential for GenAI and live streaming applications.

The Role of the Human Visual System (HVS)

A core challenge in PVQA is modeling the biological nuances of the Human Visual System. Unlike a camera sensor, the HVS is highly adaptive. For instance, "masking" occurs when the eye ignores noise in high-texture areas, and sensitivity varies based on color and contrast. Modern PVQA models increasingly incorporate these HVS characteristics—such as sensitivity, color perception, and adaptation—to improve their predictive accuracy.

Emerging Frontiers: GenAI and Immersive Media

The rise of Generative AI has shifted the definition of "quality." In GenAI, assessment is not just about technical fidelity but also about semantic consistency (alignment with text prompts) and authenticity (avoiding the "uncanny valley"). Similarly, immersive media like 360-degree video and 3D Gaussian Splatting introduce complexities such as viewport-dependent quality and motion-induced sickness, which traditional 2D metrics are unequipped to handle.

Conclusion and Future Outlook

While Deep Learning and Foundation Models (like CLIP) have revolutionized the field, significant hurdles remain. Future research must address the high computational costs of Transformer-based models and the "black box" nature of neural networks. The ultimate goal is a universal foundation model capable of assessing any visual medium—from static images to complex 3D scenes—with human-like intuition.

Summary updated

Abstract

As multimedia services such as video streaming, video conferencing, virtual reality (VR), and online gaming continue to expand, ensuring high perceptual visual quality becomes a priority to maintain user satisfaction and competitiveness. However, multimedia content undergoes various distortions during acquisition, compression, transmission, and storage, resulting in the degradation of experienced quality. Thus, perceptual visual quality assessment (PVQA), which focuses on evaluating the quality of multimedia content based on human perception, is essential for optimizing user experiences in advanced communication systems. Several challenges are involved in the PVQA process, including diverse characteristics of multimedia content such as image, video, VR, point cloud, mesh, multimodality, etc., and complex distortion scenarios as well as viewing conditions. In this paper, we first present an overview of PVQA principles and methods. This includes both subjective methods, where users directly rate their experiences, and objective methods, where algorithms predict human perception based on measurable factors such as bitrate, frame rate, and compression levels. Based on the basics of PVQA, quality predictors for different multimedia data are then introduced. In addition to traditional images and videos, immersive multimedia and generative artificial intelligence (GenAI) content are also discussed. Finally, the paper concludes with a discussion on the future directions of PVQA research.
BibTeX
@misc{zhou2025perceptual,
  author = {Wei Zhou and Hadi Amirpour and Christian Timmerer and Guangtao Zhai and Patrick Le Callet and Alan C. Bovik},
  title = {{Perceptual Visual Quality Assessment: Principles, Methods, and Future Directions}},
  year = {2025},
  eprint = {2503.00625},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2503.00625},
}
Figure