Confusing Image Quality Assessment: Toward Better Augmented Reality Experience.

Area of Science:

• Computer Vision
• Multimedia Technology
• Human-Computer Interaction

Background:

• Augmented Reality (AR) integrates digital content with real-world environments, impacting user Quality of Experience (QoE).
• Evaluating the perceptual quality of superimposed images in AR is crucial but underexplored.
• Existing image quality assessment methods may not adequately address the unique challenges of AR.

Purpose of the Study:

• To propose a theoretical framework for evaluating the perceptual quality of superimposed images, termed confusing image quality assessment (CFIQA).
• To develop novel databases and models for CFIQA and AR image quality (ARIQA).
• To investigate the influence of visual confusion on image quality assessment algorithms for AR.

Main Methods:

• Creation of a ConFusing Image Quality Assessment (CFIQA) database with 600 reference and 300 distorted images.
• Conducting subjective quality perception experiments to understand human perception of confusing images.
• Developing and evaluating benchmark and specifically designed CFIQA models, followed by an extended ARIQA study with a dedicated database and metrics.

Main Results:

• The proposed CFIQA model demonstrates state-of-the-art performance compared to benchmark models.
• The ARIQA model shows excellent generalization ability and achieves state-of-the-art performance in evaluating AR image quality.
• The study highlights the importance of considering visual confusion in AR image quality assessment.

Conclusions:

• Visual confusion is a key factor in assessing the perceptual quality of superimposed images in AR.
• The developed CFIQA and ARIQA databases, models, and metrics provide valuable resources for AR research and development.
• The findings contribute to improving the Quality of Experience (QoE) in AR applications by enabling more accurate perceptual quality evaluation.