Attention driven foveated video quality assessment

Area of Science:

• Visual Perception
• Image Processing
• Human-Computer Interaction

Background:

• Human contrast sensitivity is influenced by foveation and attention.
• Current foveation-based video quality assessment models often overlook dynamic attention mechanisms.

Purpose of the Study:

• To develop an advanced foveal imaging model integrating visual attention for video quality assessment.
• To propose a novel fixation prediction approach simulating eye movement for dynamic foveation.
• To introduce an attention-driven foveated video quality (AFViQ) metric.

Main Methods:

• An advanced foveal imaging model incorporating visual attention was developed.
• A novel approach to predict video fixations was proposed, mimicking eye movement.
• An attention-driven contrast sensitivity function was modeled and integrated into a wavelet-based distortion visibility measure.

Main Results:

• The proposed model accurately simulates dynamic foveation by integrating visual attention.
• The novel fixation prediction approach effectively mimics eye movement patterns.
• The developed AFViQ metric demonstrates promising performance in video quality assessment.

Conclusions:

• The study successfully integrated visual attention into foveation for video quality assessment.
• The proposed AFViQ metric offers a more perceptually relevant approach to video quality evaluation.
• The findings highlight the importance of incorporating dynamic visual mechanisms in objective video quality metrics.