Objective Image Quality Optimization in Augmented Reality Using Spatial Frequency Domain Models.

Area of Science:

• Optics and Vision Science
• Human-Computer Interaction
• Image Processing

Background:

• Augmented reality (AR) systems merge digital information with the physical environment using head-mounted displays (HMDs).
• Image quality in AR HMDs is degraded by contrast reduction and noise, impacting user perception.
• Assessing and optimizing AR image quality is crucial for reliable human performance.

Purpose of the Study:

• To evaluate image quality and human perceptual performance in AR systems.
• To develop and validate a target detection model for AR imaging tasks.
• To propose an image quality optimization scheme for AR display configurations.

Main Methods:

• Conducted human and model observer studies for target detection in digital and physical AR environments.
• Developed a target detection model incorporating the complete AR system, including optical see-through.
• Compared observer model performance (spatial frequency domain) with human results using Area Under the Curve (AUC).

Main Results:

• The non-prewhitening model with eye filter closely matched human performance, especially in high noise.
• AR HMD non-uniformity degraded low-contrast target detection (below 0.02) in low noise.
• Target detectability in the physical world decreased due to contrast reduction from overlaid AR images (AUC < 0.87).

Conclusions:

• AR HMDs significantly impact image quality and perceptual performance.
• A validated image quality optimization scheme can enhance AR display configurations for improved target detection.
• The proposed method aids in optimizing AR systems for both digital and physical world imaging tasks.