Quality Index for Stereoscopic Images by Separately Evaluating Adding and Subtracting.

Area of Science:

• Computer Vision
• Image Processing
• Human Visual System

Background:

• The human visual system (HVS) is crucial for stereo image quality perception.
• Leveraging visual perception knowledge in image quality assessment (IQA) models is of significant interest.

Purpose of the Study:

• To propose a full-reference metric for stereoscopic image quality assessment (SIQA).
• To utilize binocular summation and difference channels for enhanced SIQA.

Main Methods:

• Computed binocular summation and difference maps from stereo image pairs.
• Decoupled binocular summation into additive impairments and detail losses.
• Integrated Contrast Sensitivity Function (CSF) and multi-scale structural similarity (MS-SSIM) for quality computation.
• Combined binocular summation and difference qualities into a final index.

Main Results:

• The proposed metric demonstrates high consistency with subjective quality assessments.
• The metric is shown to be a robust measure for stereoscopic image quality.
• Experimental results indirectly support the existence of distinct binocular summation and difference channels.

Conclusions:

• The developed metric effectively assesses stereoscopic image quality by modeling human visual perception.
• The study validates the importance of binocular summation and difference channels in SIQA.
• The proposed method offers a robust and perceptually relevant approach to SIQA.