No-reference stereoscopic image quality assessment based on hue summation-difference mapping image and binocular...

Area of Science:

• Computer Vision
• Image Processing
• Human Visual System

Background:

• No-reference (NR) quality assessment for stereoscopic images is crucial for 3D technology but presents significant challenges.
• Existing methods often struggle with accurately evaluating 3D image quality without a reference image.

Purpose of the Study:

• To propose a novel NR stereo image quality assessment (SIQA) method.
• To enhance the accuracy and reliability of 3D image quality evaluation.

Main Methods:

• Mimics human visual system's summation and difference channels, considering binocular interactive perception.
• Calculates summation and difference images using contrast of hue and luminance in color patches.
• Integrates visual information using filtered data from left and right viewpoints to create a summation-difference mapping image (SDMI).
• Extracts features including energy entropy, joint distribution of SDMI and depth map subband coefficients, and local log-Euclidean multivariate Gaussian descriptors.
• Employs support vector regression for quality prediction.

Main Results:

• The proposed NR-SIQA algorithm demonstrates high consistency with subjective quality assessments.
• Validation was performed across four established SIQA databases, confirming the method's robustness.

Conclusions:

• The novel NR-SIQA method effectively predicts stereoscopic image quality.
• The approach, based on human visual perception and advanced feature extraction, offers a reliable solution for 3D image quality assessment.