No-reference stereoscopic image quality assessment guided by visual hierarchical structure and binocular effects

Area of Science:

• Computer Vision
• Image Processing
• Human-Computer Interaction

Background:

• Stereoscopic image quality assessment (SIQA) is crucial for 3D systems but is complex due to the human visual system (HVS) and binocular effects.
• Existing SIQA methods struggle with computational complexity and accurately modeling binocular vision challenges like rivalry and discomfort.

Purpose of the Study:

• To develop a novel, efficient, and robust SIQA method.
• To address the limitations of current SIQA techniques by incorporating visual quality perception features and binocular effects.
• To enable practical, reference-free SIQA for 3D applications.

Main Methods:

• Proposed a new SIQA method extracting image quality-aware features linked to cortical visual perception.
• Integrated considerations for binocular effects such as rivalry and visual discomfort.
• Developed a no-reference image quality assessment approach.

Main Results:

• The proposed method achieves high prediction accuracy with moderate computational complexity.
• Effectively accounts for binocular effects in quality assessment.
• Demonstrates superior performance, effectiveness, and robustness compared to state-of-the-art methods.

Conclusions:

• The novel SIQA method offers a practical and accurate solution for 3D image and video quality evaluation.
• It overcomes the limitations of traditional methods by focusing on perceptual features and binocular vision.
• The no-reference approach broadens its applicability in real-world 3D systems.